We developed a series of lightweight models, which named `PicoDet`. Because of its excellent performance, it is very suitable for deployment on mobile or CPU.
We developed a series of lightweight models, which named `PP-PicoDet`. Because of its excellent performance, it is very suitable for deployment on mobile or CPU.
- 🌟 Higher mAP: the **first** object detectors that surpass mAP(0.5:0.95) **30+** within 1M parameters when the input size is 416.
- 🚀 Faster latency: 129FPS on mobile ARM CPU.
- 🚀 Faster latency: 150FPS on mobile ARM CPU.
- 😊 Deploy friendly: support PaddleLite/MNN/NCNN/OpenVINO and provide C++/Python/Android implementation.
- 😍 Advanced algorithm: use the most advanced algorithms and innovate, such as ESNet, CSP-PAN, SimOTA with VFL, etc.
...
...
@@ -17,40 +17,35 @@ We developed a series of lightweight models, which named `PicoDet`. Because of i
-<aname="latency">Latency:</a> All our models test on `Qualcomm Snapdragon 865(4\*A77+4\*A55)` with 4 threads by arm8 and with FP16. In the above table, test latency on [NCNN](https://github.com/Tencent/ncnn) and `Lite`->[Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite). And testing latency with code:[MobileDetBenchmark](https://github.com/JiweiMaster/MobileDetBenchmark).
- PicoDet is trained on COCO train2017 dataset and evaluated on COCO val2017.
- PicoDet used 4 or 8 GPUs for training and all checkpoints are trained with default settings and hyperparameters.
