Improve YOLO here

[Stinky-Tofu]

If you have any ideas for improving yolo, you can come up here.

[tonyma666]

excellent job! I want to digest your code before graduation so that I can find a satisfactory job! and the interpretation in your code is Chinese!!hahahaha.舒服了

[wizholy]

https://arxiv.org/abs/1903.00621

[Stinky-Tofu]

@wizholy Thanks!

[YuDamon]

Gradient Harmonized Single-stage Detector https://arxiv.org/pdf/1811.05181v1.pdf

[Stinky-Tofu]

@YuDamon OK, thanks!

[fourth-archive]

I recommend this YOLOv3 tutorial: https://github.com/ultralytics/yolov3/wiki/Train-Custom-Data

The accompanying repository works on MacOS, Windows and Linux, includes multigpu and multithreading, performs inference on images, videos, webcams, and an iOS app. It also tests to slightly higher mAPs than darknet, including on the latest YOLOv3-SPP.weights (60.7 COCO mAP), and offers the ability to train custom datasets from scratch to darknet performance, all using PyTorch :) https://github.com/ultralytics/yolov3

[basser2]

what is the difference between YOLOV3-MobilenetV2 and YOLOV3-lite?

[Stinky-Tofu]

@basser2 YOLOV3-Lite's FPN structure is lighter

[basser2]

@basser2 YOLOV3-Lite's FPN structure is lighter

ok ,how about try yoloV3+shuffleNetV2？compare with tiny-yoloV3?

[Stinky-Tofu]

@basser2 Replace backbone with shuffleNetV2......

[basser2]

@basser2 Replace backbone with shuffleNetV2..... yes,just like YOLOV3-MobilenetV2
https://github.com/opconty/face_detection_in_realtime

[Stinky-Tofu]

@basser2 It's very easy...... you can try it.

[dkashkin]

@Stinky-Tofu can you please add the size of models to the table that explains V1/V2/V3? I am particularly curious if V3 is lite enough to run on mobile devices...