NanoDet-Plus v1.0.0

A stable version of NanoDet-Plus with PyTorch 1.x.

It requires pytorch-lightning>=1.9.0,<2.0.0 and torch>=1.10,<2.0.

What's Changed

• [Enhance]: Disable DDP when using a single GPU. by @RangiLyu in https://github.com/RangiLyu/nanodet/pull/369
• update version and readme by @RangiLyu in https://github.com/RangiLyu/nanodet/pull/370
• Update requirements.txt by @wwdok in https://github.com/RangiLyu/nanodet/pull/391
• [Feature]: Support print per class AP. by @RangiLyu in https://github.com/RangiLyu/nanodet/pull/395
• fix typo in README.md by @tuduweb in https://github.com/RangiLyu/nanodet/pull/397
• DEBUG: list type can not match cv2.mat by @Shawn-Tao in https://github.com/RangiLyu/nanodet/pull/423
• upgrade lightning version by @RangiLyu in https://github.com/RangiLyu/nanodet/pull/458
• Update README.md by @tpoisonooo in https://github.com/RangiLyu/nanodet/pull/438
• Fix demo libtorch error by @jedi007 in https://github.com/RangiLyu/nanodet/pull/420
• [Feature] Support timm backbones. by @RangiLyu in https://github.com/RangiLyu/nanodet/pull/399
• Add mem monitor and fix strategy by @RangiLyu in https://github.com/RangiLyu/nanodet/pull/462
• [Feature] Support tuning parameter-level optim hyperparameters. by @RangiLyu in https://github.com/RangiLyu/nanodet/pull/463
• [Feature] Support ignore boxes in nanodet head by @zero0kiriyu in https://github.com/RangiLyu/nanodet/pull/480
• Fixes a couple of issues to add fp16 training support by @RangiLyu in https://github.com/RangiLyu/nanodet/pull/488
• support lightning-1.9.0 by @RangiLyu in https://github.com/RangiLyu/nanodet/pull/489
• Implemented yolo dataset support by @cansik in https://github.com/RangiLyu/nanodet/pull/487
• bump version to 1.0.0 by @RangiLyu in https://github.com/RangiLyu/nanodet/pull/502

New Contributors

• @tuduweb made their first contribution in https://github.com/RangiLyu/nanodet/pull/397
• @Shawn-Tao made their first contribution in https://github.com/RangiLyu/nanodet/pull/423
• @tpoisonooo made their first contribution in https://github.com/RangiLyu/nanodet/pull/438
• @jedi007 made their first contribution in https://github.com/RangiLyu/nanodet/pull/420
• @zero0kiriyu made their first contribution in https://github.com/RangiLyu/nanodet/pull/480
• @cansik made their first contribution in https://github.com/RangiLyu/nanodet/pull/487

Full Changelog: https://github.com/RangiLyu/nanodet/compare/v0.4.2...v1.0.0

NanoDet-Plus v1.0.0-alpha

In NanoDet-Plus, we propose a novel label assignment strategy with a simple assign guidance module (AGM) and a dynamic soft label assigner (DSLA) to solve the optimal label assignment problem in lightweight model training. We also introduce a light feature pyramid called Ghost-PAN to enhance multi-layer feature fusion. These improvements boost previous NanoDet's detection accuracy by 7 mAP on COCO dataset.

Model checkpoints and weights

Download in the release files.

v0.4.2

Fix some compatibility issue of NanoDet v0.4

Fix pytorch-lightning compatibility. (#304 #309 ) Fix pytorch1.9 compatibility. (#308 ) Support not raising an error when evaluate with empty results. (#310)

I'm doing a lot of refactoring. NanoDet v1.x is coming soon.

Download pretrained models

v0.4.1

This is a final release of NanoDet v0.x.

I'm doing a lot of refactoring. NanoDet v1.x is coming soon.

Download pretrained models

What's new in v0.4.0

1. Fix a little bug in demo.py by BlainWu (#210)
2. Add script to export TorchScript model by strawberrypie (#211)
3. Use fixed output names when exporting ONNX (#218)
4. Use scale_factor instead of fixed size in resize to support dynamic shape inference (#218)
5. Ensure num_classes equal len(class_names) by ZHEQIUSHUI (#221)
6. Fix a bug in mnn demo while using GPU device by AcherStyx (#234)
7. Fix with_last_conv bug in shufflenet (#239)
8. Support batch eval (#241)
9. Add nanodet-m-1.5x models (#242)
10. Update model benchmark (#246)
11. Prevent lightning Trainer from disabling cudnn.benchmark (#249)
12. Fix multi-GPU evaluation bug with pytorch-lightning (#254)

Download pretrained models

Download ncnn models below

What's new in v0.3.0

1. Refactor training and testing code with pytorch-lightning.
2. Solving ONNX inference AxisError by zshn25 (#198).

Download pretrained models

What's new in v0.2.0

1. Add pyncnn demo by caishanli (#167).
2. Fix ncnn demo build failure without vulkan by nihui (#168).
3. Add NanoDet-t with Transformer Attention Network (#183).
4. Add Notebook demo by zhiqwang (#188).
5. Add feature of saving demo inference result by wwdok (#191).
6. Fix utf-8 decode bug (#184).
7. Fix test bug.

Download pretrained models

What's new in v0.1.0

1. Support MNN python and cpp inference (#83 ).
2. Support OpenVINO inference.
3. Support libtorch inference experimentally.
4. Add NanoDet-g.
5. Add EfficientNet-Lite and Rep-VGG backbone.
6. Add Model Zoo and provide more pre-trained model.
7. Refactor GFL head (#154 ).

Download pretrained models

NanoDet ncnn model released.