Image Analysis Task Segmentation for Crash-Induced Frame Loss
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image analysis systems suffer from long-time frame loss during video playback due to program crashes on computing nodes, leading to unreliability as entire image analysis tasks are discarded, resulting in significant data loss.
Innovation Solution
The central node splits image analysis tasks into sub-tasks upon a program crash, sequentially transmits these sub-tasks to other computing nodes, and discards only the error sub-tasks, retaining other images, thus minimizing data loss and improving reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the entire image analysis task is discarded upon program crash, then the computing node can be restored to proceed to the next task, but the entire task's images are lost resulting in long-time frame loss during video playback
Solution Approach 1:
The patent divides an image analysis task into multiple independent image analysis sub-tasks, each processing a single image. When a program crash occurs, only the failed sub-task is discarded rather than the entire task. This segmentation allows the system to retain and reprocess successful sub-tasks, significantly reducing image loss and frame loss duration during video playback.
2Ease of repair
If the entire image analysis task is discarded upon program crash, then the computing node can be restored quickly, but the duration of frame loss during video playback increases
Solution Approach 1:
By segmenting the task into sub-tasks, the system can quickly restore the computing node after a crash while minimizing frame loss. Only the failed sub-task needs to be reprocessed, allowing the majority of the video sequence to be recovered without prolonged frame loss.
3Productivity
If image analysis tasks are processed in parallel on multiple computing nodes, then processing efficiency is improved, but program crashes can cause entire task loss reducing reliability
Solution Approach 1:
The patent maintains parallel processing efficiency by dividing tasks into sub-tasks that can be distributed across multiple computing nodes. Simultaneously, it improves reliability by ensuring that a crash on one node only affects individual sub-tasks rather than entire tasks, allowing other parallel operations to continue successfully.
Solution Approach 2:
The patent implements a selective discard and recover mechanism where only failed sub-tasks are discarded upon crash, while successful sub-tasks are retained and reprocessed. This minimizes data loss and maintains high reliability in parallel processing environments.
Data Source
Figure 1~2
Figure 3
Figure 4-1~4-2
AI summary
The present application discloses an image analysis method and apparatus, upon a program crash on a computing node, the central node may acquire the image analysis task that results in the crash, split the image analysis task into multiple image analysis sub-tasks, transmit sequentially the image analysis sub-tasks to the computing node until a program crash occurs thereon. The central node then acquires an error image analysis sub-task that results in the crash on the computing node, and discard the error image analysis sub-task. Compared with relevant art, the image analysis method according to the embodiment of the present application discards only the crash resulting image from an image analysis task, while other images in the image analysis task are retained. As such, fewer images are lost, so that the duration of a frame loss during video playback can be reduced and the reliability of image analysis can be improved. The approach is applicable for image analysis.