Method for quality control of contact lenses

Abstract

A method for quality control of contact lenses is disclosed herein. An exemplary method includes receiving an input image showing a target contact lens. The exemplary method includes outputting a first foreign object metric and non-conformance data based on the analysis of the input image and using a first learning model. The exemplary method includes outputting an artifact metric using an artifact model based at least on the non-conformance data. The artifact model may be implemented based on one or more artifact attributes. The exemplary method includes outputting a second foreign object metric using a second learning model based at least on the first foreign object metric and the artifact metric. The second learning model may be scale-dependent. The second foreign object metric indicates at least the pass or fail status of the target contact lens.

Description

Background Art

[0001] Quality control of contact lenses may rely on inspection and detection techniques. However, improvements are needed.

Summary of the Invention

Means for Solving the Problems

[0002] A method for quality control of contact lenses and / or contact lens packages is disclosed herein. An exemplary method may include receiving an input image showing a target contact lens. An exemplary method may include analyzing the input image using a first learning model. The first learning model may not be scale-dependent. An exemplary method may include outputting a first foreign object metric and non-conformance data based on the analysis of the input image. An exemplary method may include using an artifact model to identify false non-conformance data and true non-conformance data from the non-conformance data. The artifact model may be implemented based on one or more artifact attributes. The false non-conformance data may indicate the presence of one or more artifact attributes. An exemplary method may include outputting an artifact metric based on the false non-conformance data and one or more of the false non-conformance data. An exemplary method may include using a second learning model to output a second foreign object metric based on at least the first foreign object metric and the artifact metric. The second learning model may be scale-dependent. The second foreign object metric may indicate at least the pass or fail status of the target contact lens.

[0003] The first learning model may be trained and tested on a plurality of images including pass or fail status. The artifact model may be trained and tested on images including at least one or more artifact attributes. The one or more artifact attributes may include transparent bubbles, folded lenses, or no artifacts, or combinations thereof.

[0004] The second learning model can be trained and tested on a plurality of images including at least a pass or fail state. The second learning model can be configured to recover false fails based on the scale of the defect that caused the fail. The second learning model can be configured to recover false fails based on suspected artifacts that are less than a predetermined size.

[0005] A method for quality control of contact lenses is disclosed herein. An exemplary method can include receiving an input image showing a target contact lens. An exemplary method can include outputting a first foreign object metric and fail data based on the analysis of the input image and using a first learning model. An exemplary method can include outputting an artifact metric using an artifact model based on at least the fail data. The artifact model can be implemented based on one or more artifact attributes. An exemplary method can include outputting a second foreign object metric using a second learning model based on at least the first foreign object metric and the artifact metric. The second learning model can be scale-dependent. The second foreign object metric can indicate at least a pass or fail state of the target contact lens.

[0006] The first learning model can be trained and tested on a plurality of images including a pass or fail state. The artifact model can be trained and tested on images including at least one or more artifact attributes. The one or more artifact attributes can include a transparent bubble, a folded lens, or no artifact, or a combination thereof.

[0007] The second learning model can be trained and tested on a plurality of images including at least a pass or fail state. The second learning model can be configured to recover false fails based on the scale of the defect that caused the fail. The second learning model can be configured to recover false fails based on suspected artifacts that are less than a predetermined size. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The following drawings generally illustrate various embodiments contemplated in the present disclosure by way of example and not limitation. In the drawings, the following applies.

Figure 1

Figure 2

DETAILED DESCRIPTION OF THE INVENTION

[0009] A method for quality control of a contact lens and / or a contact lens package is disclosed herein. The lens package may require quality control to prevent foreign objects from entering the package. The package may also require inspection to identify holes in the package or lens, as well as edge defects in the lens. Foreign objects can damage or contaminate the lens, both of which are disadvantageous in meeting customer needs. The package can be inspected at any stage of the process, including before and after the package is sealed.

[0010] In the following detailed description, reference is made to the accompanying drawings which form a part hereof and which illustrate by way of example specific embodiments in which the invention may be practiced. It is to be understood that other embodiments may be utilized and structural changes may be made without departing from the scope of the present invention. It is understood that electrical, mechanical, logical, and structural changes may be made to the embodiments without departing from the spirit and scope of the present teachings. Accordingly, the following detailed description should not be construed in a limiting sense, and the scope of the present disclosure is defined by the appended claims and their equivalents.

[0011] In-line model - The in-line model may include a conventional deep learning model and an in-line ensemble model, and is used to reduce false rejects and support maintaining high capabilities. As an exemplary example, an in-line model may be used. First, the ensemble model may include a highly capable deep learning model combined with an artifact model. The artifact model may be used to recover false rejects due to folded lenses and bubbles. A second deep learning model that is in-line with the ensemble model may be used to recover false rejects due to suspicious artifacts smaller than a predetermined size. The first deep learning model may be scale-dependent, while the second deep learning model is scale-independent. The ordering of these models, first scale-independent and then scale-dependent, enables high capabilities by the first model (capturing suspicious artifacts regardless of size) and then reduction of false rejects (due to size) by the second model. If the order of these models is reversed, the pool of all detected true rejects will be smaller due to scale-dependence, and the overall in-line model capabilities will be lower. FIG. 1 shows the ordering of ML models forming an exemplary in-line model according to one aspect of the present disclosure.

[0012] Ensemble model - The ensemble model combines a standard deep learning model (support vector machine) in cooperation with an artifact model. These models are executed in parallel and may be incorporated into the ensemble model to generate an optimal model having high detection capabilities and low false rejects (see FIG. 2).

[0013] As an exemplary example, the ensemble was trained with 3018 images (see Table 1 for details).

[0014]

Table 1

[0015] In this case, the purpose of the artifact model is twofold. First, to recover false rejections from the deep learning model 1. Second, to strengthen true rejections from the deep learning model 1. The artifact model was trained using 833 images covering four main attributes as described in Table 2 below. Other artifact attributes may be used.

[0016]

Table 2

[0017] A method for quality control of contact lenses is disclosed herein. The method may be executed by a computing device. Instructions for performing the method steps may be stored in the memory of the computing device. The method steps may be executed by a processor of the computing device.

[0018] An input image showing the target contact lens may be received.

[0019] The input image may be analyzed using a first learning model. The first learning model may not be scale-dependent. The first learning model may comprise a deep learning model. The first learning model may be trained and tested on a plurality of images including pass or fail states.

[0020] A first foreign object metric and fail data may be output based on the analysis of the input image.

[0021] False rejection data and true rejection data may be identified using an artifact model from the fail data. The artifact model may be implemented based on one or more artifact attributes. The false rejection data may indicate the presence of one or more artifact attributes. The artifact model may be trained and tested on images including at least one or more artifact attributes. The one or more artifact attributes may include transparent bubbles, folded lenses, or no artifacts, or combinations thereof.

[0022] The artifact metric can be output based on one or more of the false non - conforming data and the false non - conforming data.

[0023] The second foreign object metric can be output based on at least the first foreign object metric and the artifact metric using a second learning model. The second learning model can be scale - dependent. The second foreign object metric can indicate at least the pass or fail status of the target contact lens. The second learning model can comprise a deep learning model. The second learning model can be trained and tested on a plurality of images including at least the pass or fail status. The second learning model can be configured to recover false non - conformances based on the scale of the defect that caused the non - conformance. The second learning model can be configured to recover false non - conformances based on suspected artifacts that are less than a predetermined size.

[0024] A method for quality control of contact lenses is disclosed herein. The method can be executed by a computing device. Instructions for performing the method steps can be stored in the memory of the computing device. The method steps can be executed by a processor of the computing device.

[0025] An input image representing a target contact lens can be received.

[0026] Based on the analysis of the input image and using a first learning model, a first foreign object metric and non - conforming data based on the analysis of the input image can be output. The first learning model can comprise a deep learning model. The first learning model can be trained and tested on a plurality of images including the pass or fail status.

[0027] An artifact metric can be output based at least on nonconforming data using an artifact model. The artifact model can be implemented based on one or more artifact attributes. The artifact model can be trained and tested on an image including at least one or more artifact attributes. The one or more artifact attributes can include a transparent bubble, a folded lens, or no artifact, or a combination thereof.

[0028] A second foreign object metric can be output based at least on the first foreign object metric and the artifact metric using a second learning model. The second learning model can be scale-dependent. The second foreign object metric can indicate at least a pass or fail state of the target contact lens. The second learning model can comprise a deep learning model. The second learning model can be trained and tested on a plurality of images including at least a pass or fail state. The second learning model can be configured to recover a false fail based on the scale of a defect that caused the fail. The second learning model can be configured to recover a false fail based on a suspected artifact that is less than a predetermined size. Any desired size threshold can be used, such as less than 1000 microns, less than 900 microns, less than 800 microns, less than 700 microns, less than 600 microns, less than 500 microns, less than 400 microns, 300 microns, less than 200 microns, or less than 100 microns.

[0029] Aspect The present disclosure includes at least the following aspects.

[0030] Aspect 1: A method for quality control of contact lenses, comprising receiving an input image showing a target contact lens, analyzing the input image using a scale-independent first learning model, outputting a first foreign object metric and non-conformance data based on the analysis of the input image, using an artifact model implemented based on one or more artifact attributes to identify false non-conformance data and true non-conformance data from the non-conformance data, wherein the false non-conformance data indicates the presence of one or more artifact attributes, identifying, outputting an artifact metric based on the false non-conformance data and one or more of the false non-conformance data, and using a second learning model to output a second foreign object metric based on at least the first foreign object metric and the artifact metric, wherein the second learning model is scale-dependent and the second foreign object metric indicates at least the pass or fail status of the target contact lens.

[0031] Aspect 2: The method according to aspect 1, wherein the first learning model includes a deep learning model.

[0032] Aspect 3: The method according to aspect 1 or 2, wherein the first learning model is trained and tested on a plurality of images including pass or fail statuses.

[0033] Aspect 4: The method according to any one of aspects 1 to 3, wherein the artifact model is trained and tested on images including at least one or more artifact attributes.

[0034] Aspect 5: The method according to any one of aspects 1 to 4, wherein the one or more artifact attributes include transparent bubbles, folded lenses, or no artifacts, or combinations thereof.

[0035] Aspect 6: The method according to any one of aspects 1 to 5, wherein the second learning model includes a deep learning model.

[0036] Aspect 7: The method according to any one of Aspects 1 to 6, wherein the second learning model is trained and tested on a plurality of images including at least a pass or fail state.

[0037] Aspect 8: The method according to any one of Aspects 1 to 7, wherein the second learning model is configured to recover a false fail based on the scale of a defect that caused the fail.

[0038] Aspect 9: The method according to any one of Aspects 1 to 8, wherein the second learning model is configured to recover a false fail based on a suspected artifact that is less than a predetermined size.

[0039] Aspect 10: A method for quality control of contact lenses, comprising receiving an input image showing a target contact lens, and based on the analysis of the input image and using a first learning model, outputting a first foreign object metric and fail data based on the analysis of the input image, outputting an artifact metric based at least on the fail data using an artifact model implemented based on one or more artifact attributes, and outputting a second foreign object metric using a second learning model based at least on the first foreign object metric and the artifact metric, wherein the second learning model is scale-dependent, and the second foreign object metric indicates at least a pass or fail state of the target contact lens.

[0040] Aspect 11: The method according to Aspect 10, wherein the first learning model includes a deep learning model.

[0041] Aspect 12: The method according to Aspect 10 or 11, wherein the first learning model is trained and tested on a plurality of images including a pass or fail state.

[0042] Aspect 13: The method according to any one of Aspects 10 to 12, wherein the artifact model is trained and tested on an image including at least one or more artifact attributes.

[0043] Aspect 14: The method according to any one of Aspects 10 to 13, wherein one or more artifact attributes include a transparent bubble, a folded lens, or no artifact, or a combination thereof.

[0044] Aspect 15: The method according to any one of Aspects 10 to 14, wherein the second learning model includes a deep learning model.

[0045] Aspect 16: The method according to any one of Aspects 10 to 15, wherein the second learning model is trained and tested on a plurality of images including at least a qualified or unqualified state.

[0046] Aspect 17: The method according to any one of Aspects 10 to 16, wherein the second learning model is configured to recover a false non - qualification based on the scale of the defect that caused the non - qualification.

[0047] Aspect 18: The method according to any one of Aspects 10 to 17, wherein the second learning model is configured to recover a false non - qualification based on a suspected artifact that is less than a predetermined size.

[0048] Aspect 19: The method according to any one of Aspects 1 to 18, wherein the second learning model is configured to recover a false non - qualification based on a suspected artifact having a length less than 1000 microns, less than 900 microns, less than 800 microns, less than 700 microns, less than 600 microns, less than 500 microns, less than 400 microns, 300 microns, less than 200 microns, or less than 100 microns.

[0049] In this system or method, when an image is acquired, before analyzing the image with the model, pre - processing of the image can be performed. The pre - processing of the image may include identifying a specific region of interest, improving the contrast, adding or removing other acquired images, etc.

[0050] For the identification of packages to be rejected, a convolutional neural network (CNN) may be designed, modified, and implemented. The packages may be sealed or unsealed, and may contain lenses or be empty. In addition to the packages may or may not contain lenses, they may or may not contain liquid. There are multiple platforms for creating CNNs, and it will be understood by those skilled in the art that they include TensorFlow (Google), Caffe (UC Berkeley), CNTK (Microsoft), Theano (LISA lab), Keras (Google), Matlab (Mathworks), Octave & Python. Existing modules may be modified to slice the original image into smaller sizes that only contain regions of interest (ROIs) of a size appropriate for providing to the CNN. In this case, these images are defined by regions of interest (ROIs), for example 299 pixels × 299 pixels (monochrome), defined by the package itself so that each ROI provided to the CNN has a suitable number of pixels for processing. An exemplary CNN such as Inception V3 can include a total of 48 layers with a mixture of convolutional layers, max pooling layers, and fully connected layers. In a later stage of the network, a rectifying linear unit (ReLU) is used, and in addition, a significant number of nodes can be randomly removed to avoid overfitting. For example, filters of size 3×3 may be used with a stride of 1, 2, or more numbers. Layers can use padding around the edges and corners of the image to improve image processing. Similarly, various layers of the network (e.g., convolutional layers, max pooling layers, softmax layers, dropout layers, fully connected layers, etc.) can be modified to achieve better results in identifying packages with some defects without misidentifying defect-free packages as those containing defects or FM.

[0051] Many operating systems, including Linux, UNIX (registered trademark), OS / 2 (registered trademark), and Windows (registered trademark), can execute many tasks simultaneously and are called multitasking operating systems. Multitasking is the ability of an operating system to execute multiple executable files at the same time. Each executable file is executed within its own address space, which means that executable files have no way of sharing any of their memory. Therefore, no program can impair the execution of any other program running on the system. However, programs have no way of exchanging any information other than through the operating system (or by reading files stored in the file system).

[0052] Multiprocess computing is similar to multitasking computing in that the terms task and process are often used interchangeably, although some operating systems distinguish between the two. The present invention may be, or may include, a system, method, and / or computer program product at any possible level of technological detail. A computer program product can include a computer-readable storage medium having computer-readable program instructions for causing a processor to execute aspects of the present invention. The computer-readable storage medium can be a tangible device that can hold and store instructions for use by an instruction execution device.

[0053] A computer-readable storage medium may be, for example, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing, but is not limited thereto. A non-exhaustive list of more specific examples of computer-readable storage media includes portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, mechanically encoded devices such as punch cards or raised structures in grooves, and any suitable combination of the foregoing.

[0054] As used herein, a computer-readable storage medium should not be construed to be a signal per se that is transient, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., an optical pulse passing through an optical fiber cable), or an electrical signal transmitted through a bus. The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to respective computing / processing devices, or to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface of each computing / processing device receives the computer-readable program instructions from the network and transfers the computer-readable program instructions for storage in a computer-readable storage medium within each computing / processing device.

[0055] Computer-readable program instructions for executing the operation of the present invention may be source code or object code written in any combination of one or more programming languages, including assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, integrated circuit configuration data, or object-oriented programming languages such as Smalltalk, C++, and procedural programming languages such as the "C" programming language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0056] In some embodiments, for example, an electronic circuit including a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA) may execute the computer-readable program instructions by utilizing the state information of the computer-readable program instructions to customize the electronic circuit in order to execute aspects of the present invention.

[0057] Aspects of the present invention are described herein with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It will be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer-readable program instructions. These computer-readable program instructions are provided to a processor of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus to create a machine such that the instructions are executed via the processor of the computer or other programmable data processing apparatus to implement the functions / operations specified in one or more blocks of the flowchart and / or block diagram.

[0058] These computer-readable program instructions may also be stored in a computer-readable storage medium that can cause a computer, programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer-readable storage medium storing the instructions includes a manufactured article including instructions for implementing aspects of the functions / acts specified in one or more blocks of the flowchart and / or block diagram. Loading these computer-readable program instructions onto a computer, other programmable data processing apparatus, or other device causes a series of operational steps to be performed on that computer, other programmable apparatus, or other device to generate a computer-implemented process, such that the instructions executed on the computer, other programmable apparatus, or other device implement the functions / operations defined in the blocks of the flowchart and / or block diagram. The flowcharts and block diagrams in the figures illustrate the structure, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, segment, or portion of instructions that includes one or more executable instructions for implementing the indicated logical function.

[0059] In some alternative implementations, the functions shown in the blocks may be performed in an order different from that shown in the figures. For example, two blocks shown consecutively may actually be executed substantially simultaneously or in the reverse order, depending on the relevant functions. Also note that each block of the block diagram and / or flowchart, as well as combinations of blocks of the block diagram and / or flowchart, may be implemented by a dedicated hardware-based system that performs a particular function or operation, or a combination of dedicated hardware and computer instructions. Although specific embodiments of the present invention have been described, those skilled in the art will understand that there are other embodiments equivalent to the described embodiments. Therefore, it should be understood that the present invention is not limited by the specific exemplary embodiments shown, but only by the appended claims.

[0060] From the above description, it can be seen that the present invention provides a system, a computer program product, and a method for efficiently executing the described technology. References to elements in the singular in the claims do not mean "only one" unless explicitly stated otherwise, but rather "one or more." All structural and functional equivalents of the elements of the above exemplary embodiments, whether now known or later to be known to those skilled in the art, are intended to be included within the scope of the present claims. Elements of the claims of this specification should not be construed under the provisions of paragraph 6 of section 112 of the United States Patent Act unless the element is explicitly recited using the phrase "means for" or "step for."

[0061] The foregoing written description of the invention enables one of ordinary skill in the art to make and use what is currently considered to be its best mode, but one of ordinary skill in the art will recognize the existence of alternatives, adaptations, variations, combinations, and equivalents to the specific embodiments, methods, and examples herein. One of ordinary skill in the art will understand that the present disclosure is merely illustrative and that various modifications can be made within the scope of the invention. Additionally, although a particular feature of the present teachings may be disclosed with respect to only one of several implementations, such a feature may be combined with one or more other features of one or more other implementations as desired and advantageous for any given function or particular function. Further, to the extent that the terms "including", "includes", "having", "has", "with", or variations thereof are used in either the "Description of Embodiments" or the "Claims", such terms are intended to be inclusive in a manner similar to the term "comprising".

[0062] Other embodiments of the present teachings will be apparent to one of ordinary skill in the art in view of the specification or practice of the teachings disclosed herein. Accordingly, the invention should not be limited by the described embodiments, methods, and examples, but rather should be limited by all embodiments and methods within the scope and spirit of the invention. Accordingly, the invention is not limited to the specific embodiments shown herein but is only limited by the following claims.

[0063] [Embodiments] (1) A method for quality control of contact lenses, comprising: receiving an input image showing a target contact lens; analyzing the input image using a first scale-independent learning model; outputting a first foreign object metric and non-conforming data based on the analysis of the input image; Identifying false non-conforming data and true non-conforming data from the non-conforming data using an artifact model implemented based on one or more artifact attributes, wherein the false non-conforming data indicates the presence of the one or more artifact attributes, and the identifying; Outputting an artifact metric based on one or more of the false non-conforming data and the false non-conforming data; Using a second learning model to output a second foreign object metric based on at least the first foreign object metric and the artifact metric; including; the second learning model is scale-dependent; the second foreign object metric indicates at least the pass or fail state of the target contact lens. (2) The method according to Embodiment 1, wherein the first learning model includes a deep learning model. (3) The method according to Embodiment 1, wherein the first learning model is trained and tested on a plurality of images including pass or fail states. (4) The method according to Embodiment 1, wherein the artifact model is trained and tested on an image including at least the one or more artifact attributes. (5) The method according to Embodiment 1, wherein the one or more artifact attributes include a transparent bubble, a folded lens, or no artifact, or a combination thereof.

[0064] (6) The method according to Embodiment 1, wherein the second learning model includes a deep learning model. (7) The method according to Embodiment 1, wherein the second learning model is trained and tested on a plurality of images including at least pass or fail states. (8) The method according to Embodiment 1, wherein the second learning model is configured to recover false non-conformities based on the scale of the defect causing the non-conformity. (9) The method according to Embodiment 1, wherein the second learning model is configured to recover a false rejection based on a suspicious artifact smaller than a predetermined size. (10) A method for quality control of contact lenses, comprising: receiving an input image showing a target contact lens; outputting a first foreign object metric and rejection data based on the analysis of the input image using a first learning model that is independent of scale based on the analysis of the input image; outputting an artifact metric based at least on the rejection data using an artifact model implemented based on one or more artifact attributes; outputting a second foreign object metric based at least on the first foreign object metric and the artifact metric using a second learning model; comprising: wherein the second learning model is scale-dependent; wherein the second foreign object metric indicates at least a pass or fail state of the target contact lens.

[0065] (11) The method according to Embodiment 10, wherein the first learning model is trained and tested on a plurality of images including pass or fail states. (12) The method according to Embodiment 10, wherein the artifact model is trained and tested on an image including at least the one or more artifact attributes. (13) The method according to Embodiment 10, wherein the one or more artifact attributes include a transparent bubble, a folded lens, or no artifact, or a combination thereof. (14) The method according to Embodiment 10, wherein the second learning model is trained and tested on a plurality of images including at least pass or fail states. (15) The method according to Embodiment 10, wherein the second learning model is configured to recover a false rejection based on the scale of a defect that caused the rejection.

[0066] (16) The method according to embodiment 10, wherein the second learning model is configured to recover a false rejection based on an artifact suspected of being less than a predetermined size. (17) A method for quality control of contact lenses, receiving an input image showing a target contact lens; outputting a first foreign object metric and rejection data based on the analysis of the input image using a first model, the first learning model being scale-independent; outputting an artifact metric based at least on the rejection data using an artifact model implemented based on one or more artifact attributes; outputting a second foreign object metric based at least on the first foreign object metric and the artifact metric using a second model; comprising the second model being scale-dependent; the second foreign object metric indicating at least a pass or fail state of the target contact lens. (18) The method according to embodiment 17, wherein the first learning model is trained and tested on a plurality of images including pass or fail states. (19) The method according to embodiment 17, wherein the artifact model is trained and tested on an image including at least the one or more artifact attributes. (20) The method according to embodiment 17, wherein the one or more artifact attributes include a transparent bubble, a folded lens, or no artifact, or a combination thereof.

[0067] (21) The method according to embodiment 17, wherein the second learning model is trained and tested on a plurality of images including at least pass or fail states. (22) The method according to Embodiment 17, wherein the second learning model is configured to recover a false non-conformance based on the scale of the defect that caused the non-conformance. (23) The method according to Embodiment 17, wherein the second learning model is configured to recover a false non-conformance based on a suspected artifact that is less than a predetermined size.

Claims

1. A method for quality control of contact lenses, comprising: receiving an input image showing a target contact lens; analyzing the input image using a first learning model independent of scale; outputting a first foreign object metric and non-conformance data based on the analysis of the input image; identifying false non-conformance data and true non-conformance data from the non-conformance data using an artifact model implemented based on one or more artifact attributes, wherein the false non-conformance data indicates the presence of the one or more artifact attributes; outputting an artifact metric based on the false non-conformance data and one or more of the false non-conformance data; outputting a second foreign object metric using a second learning model based on at least the first foreign object metric and the artifact metric; wherein the second learning model is scale-dependent; the second foreign object metric indicates at least the pass or fail status of the target contact lens.

2. The method according to claim 1, wherein the first learning model includes a deep learning model.

3. The method according to claim 1, wherein the first learning model is trained and tested on a plurality of images including pass or fail status.

4. The method according to claim 1, wherein the artifact model is trained and tested on an image including at least the one or more artifact attributes.

5. The method according to claim 1, wherein the one or more artifact attributes include transparent bubbles, folded lenses, or no artifacts, or combinations thereof.

6. The method according to claim 1, wherein the second learning model includes a deep learning model.

7. The method according to claim 1, wherein the second learning model is trained and tested on a plurality of images including at least pass or fail status.

8. The method according to claim 1, wherein the second learning model is configured to recover false non-conformance based on the scale of the defect causing the non-conformance.

9. The method according to claim 1, wherein the second learning model is configured to recover false non-conformance based on suspected artifacts smaller than a predetermined size.

10. A method for quality control of contact lenses, comprising: Receiving an input image showing a target contact lens; Based on the analysis of the input image and using a first learning model independent of scale, outputting a first foreign object metric and non-conforming data based on the analysis of the input image; Using an artifact model implemented based on one or more artifact attributes, outputting an artifact metric based at least on the non-conforming data; Using a second learning model, outputting a second foreign object metric based at least on the first foreign object metric and the artifact metric; comprising; the second learning model is scale-dependent; the second foreign object metric indicates at least a pass or fail status of the target contact lens.

11. The method according to claim 10, wherein the first learning model is trained and tested on a plurality of images including pass or fail status.

12. The method according to claim 10, wherein the artifact model is trained and tested on images including at least the one or more artifact attributes.

13. The method according to claim 10, wherein the one or more artifact attributes include transparent bubbles, folded lenses, or no artifacts, or combinations thereof.

14. The method according to claim 10, wherein the second learning model is trained and tested on a plurality of images including at least pass or fail status.

15. The method according to claim 10, wherein the second learning model is configured to recover false non-conformances based on the scale of the defect causing the non-conformance.

16. The method according to claim 10, wherein the second learning model is configured to recover false non-conformances based on suspected artifacts smaller than a predetermined size.

17. A method for quality control of contact lenses, comprising: Receiving an input image showing a target contact lens; Based on the analysis of the input image and using a first model, outputting a first foreign object metric and non-conforming data based on the analysis of the input image, wherein the first learning model is scale-independent. Using an artifact model implemented based on one or more artifact attributes, output an artifact metric based at least on the non-conforming data. Using a second model, output a second foreign object metric based at least on the first foreign object metric and the artifact metric. comprising the second model is scale-dependent. the second foreign object metric indicates at least the pass or fail status of the target contact lens. **Claim 18** The method according to claim 17, wherein the first learning model is trained and tested on a plurality of images including pass or fail statuses. **Claim 19** The method according to claim 17, wherein the artifact model is trained and tested on an image including at least the one or more artifact attributes. **Claim 20** The method according to claim 17, wherein the one or more artifact attributes include a transparent bubble, a folded lens, or no artifact, or a combination thereof. **Claim 21** The method according to claim 17, wherein the second learning model is trained and tested on a plurality of images including at least pass or fail statuses. **Claim 22** The method according to claim 17, wherein the second learning model is configured to recover a false non-conformance based on the scale of the defect causing the non-conformance. **Claim 23** The method according to claim 17, wherein the second learning model is configured to recover a false non-conformance based on a suspected artifact that is less than a predetermined size.

Images