Model training method and device, equipment and storage medium

Abstract

The embodiment of the invention discloses a model training method and device, equipment and a storage medium. The method comprises the steps: firstly acquiring a training sample set, wherein the training sample set at least comprises difficult samples of which labels are unknown categories and the unknown categories are other categories except a positive category or a negative category; performing supervised training on a to-be-trained model by using a quantized cross entropy loss function based on the training sample set, wherein the quantized cross entropy loss function comprises loss functions suitable for predicting the positive category, the negative category and the unknown category, respectively; and under the condition that the training result shows that the predicted confidence of the difficult sample is between two preset confidence thresholds, acquiring the trained target model.

Description

technical field [0001] This application relates to the field of computer vision, involving but not limited to model training methods, devices, equipment, and storage media. Background technique [0002] The image binary classification problem is a kind of basic problem in deep learning, and it is also a basic problem in computer vision. It is more commonly used in image recognition (image classification). [0003] In the use of actual scenarios, the input of the model will not only be negative samples with clear labels or positive samples with clear labels, but also samples with unclear labels may appear as the input of the model (that is, samples cannot be clearly marked as negative or positive. ). Since the model learns based on well-labeled negative samples and positive samples during the training phase, when encountering samples with unclear labels as input during actual use, the model cannot effectively predict this type of samples, and may eventually be randomly class...

AI Technical Summary

Problems solved by technology

Since the model learns based on well-labeled negative samples and positive samples during the training phase, when encountering samples with unclear labels as input during actual use, the model cannot effectively predict this type of samples, and may eventually be randomly classified into Positive or negative categories, resulting in the final negative category and the precision on the positive category being affected

Images