What are the characteristics of small objects and how to design an algorithm based on the characteristics ? To my knowledge, feature fusion and context learning are usually used to improve object detection. However, it is hard to why they improve the detection results. Are there some algorithms designed just for small object detection ?
In this Journal maybe help your problem: Small Object Detection Algorithm Based on Feature Pyramid-Enhanced Fusion SSD
Note: Small objects can be estimated using SSD object detection algorithm
Thank you for your replay. I have ever use ssd to do some experiments. However, the detection performance is still not satisfied for small objects.
Dear Jiaxu Leng
I think if you read this Quora article you will find your solution: https://www.quora.com/How-do-you-do-object-detection-using-CNNs-on-small-objects-like-ping-pong-balls
you could refer to this paper Preprint Small Object Detection using Context and Attention
Tianlin Zhang Thank you for your suggestion. Context and Attention are used in this paper and improve the final detection performance. HOWEVER, it is inexplicable that how they improve detection accuracy. Why can multi-scale feature fusion improve the feature representation? Is it possible that feature fusion may cause worse feature presentation? I'm confused!
We have done some research on the detection of sperm that are small objects and have a very good percentage. You can use our article on deep training,
https://www.researchgate.net/publication/339199542_Sperm_detection_and_tracking_in_phase-contrast_microscopy_image_sequences_using_deep_learning_and_modified_CSR-DCF/stats
i deem that the definition of small objects is of great importance for your research.
- Badges
- Science topic
- Similar topics
- Psychology
- Mental Processes
- Learning