szq0214/DSOD
A research implementation of a deeply supervised object detection model that trains from scratch without pre-training.
Not currently ranked — collecting fresh signals.
star history
DSOD implements an object detection architecture published at ICCV 2017, enabling object detection models to be trained from scratch without relying on ImageNet pre-trained weights. The method uses deep supervision techniques to improve training convergence when training detectors from scratch. The project includes model implementations and training code for object detection on standard benchmarks.
Frequently asked
- What is szq0214/DSOD?
- A research implementation of a deeply supervised object detection model that trains from scratch without pre-training.
- Is DSOD open source?
- Yes — szq0214/DSOD is an open-source project tracked on heatdrop.
- What language is DSOD written in?
- szq0214/DSOD is primarily written in Python.
- How popular is DSOD?
- szq0214/DSOD has 711 stars on GitHub.
- Where can I find DSOD?
- szq0214/DSOD is on GitHub at https://github.com/szq0214/DSOD.