Related Experiment Video
Updated: Sep 25, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
659
TOD-CNN: An effective convolutional neural network for tiny object detection in sperm videos.
Shuojia Zou1, Chen Li1, Hongzan Sun2
1Microscopic Image and Medical Image Analysis Group, College of Medicine and Biological Information Engineering, Northeastern University, Shenyang, China.
Computers in Biology and Medicine
|April 28, 2022
Summary
We developed a new AI model, the tiny object detection convolutional neural network (TOD-CNN), to accurately detect tiny objects like sperm in microscopic videos. This technology aids in precise sperm quality analysis.
Area of Science:
- Computer Vision
- Biomedical Imaging
- Artificial Intelligence
Background:
- Detecting tiny objects in microscopic videos, such as sperm, is challenging due to image quality issues like fuzziness and irregular shapes.
- Existing methods struggle with precise object localization, hindering large-scale experiments and accurate analysis.
Purpose of the Study:
- To introduce a novel convolutional neural network (CNN) for tiny object detection (TOD-CNN) specifically designed for microscopic videos.
- To address the limitations of current methods in detecting and precisely positioning small, irregular objects like sperm.
Main Methods:
- Development of a specialized CNN architecture (TOD-CNN) tailored for tiny object detection.
- Creation of a comprehensive dataset comprising 111 high-quality sperm microscopic videos with over 278,000 annotated objects.
- Design of a graphical user interface (GUI) for effective model deployment and testing.
Main Results:
- The TOD-CNN achieved a high accuracy of 85.60% AP50 in real-time sperm detection within microscopic videos.
- The model demonstrated robust performance in handling fuzzy and irregularly shaped objects.
Conclusions:
- The TOD-CNN offers a significant advancement in the automated detection of sperm in microscopic videos.
- This technology has strong potential for improving the accuracy and efficiency of sperm quality analysis and related diagnostic processes.
