📚 LINK TO BLOGPOST: learnopencv.com/yolox-object-detector-paper-explanation-and-custom-training/ ▶ LINK TO YOLO MASTERCLASS PLAYLIST: th-cam.com/play/PLfYPZalDvZDLALsG9o-cjwNelh-oW9Xc4.html
Hi, may I know that it does the back propagation through the sum of total loss function or the loss function in classification head and regression head do the back propagation separately?
Hello. I want to use YOLOX for Object detection on CPU. Will it give a good FPS? Which YOLO models is good to use on CPU? Also whatever the model is I’m supposed to use OPENCV DNN only i.e. convert yolox model from pytorch to onnx and use it with opencv. Is there any way to improve the FPS?
Hey! You can try with nano, tiny, and small models. Performance will vary from device to device. It is recommended to test with the pre-trained ONNX models before custom training. Keep input images below 720p for optimal speed.
It is faster and more accurate than CenterNet. However there are new models for CenterNet like CenterNet2 and CenterNet++. We are yet to explore them in detail.
📚 LINK TO BLOGPOST: learnopencv.com/yolox-object-detector-paper-explanation-and-custom-training/
▶ LINK TO YOLO MASTERCLASS PLAYLIST: th-cam.com/play/PLfYPZalDvZDLALsG9o-cjwNelh-oW9Xc4.html
good evening- this is super .catch you later! ;)
Your video is extremely clear and inspiring, please keep going!
Thank you, Oliver!
Hi, may I know that it does the back propagation through the sum of total loss function or the loss function in classification head and regression head do the back propagation separately?
Hello. It does a backpropagation of the total loss.
excellent summary, thankyou
Great explanation.
Hello. I want to use YOLOX for Object detection on CPU. Will it give a good FPS? Which YOLO models is good to use on CPU? Also whatever the model is I’m supposed to use OPENCV DNN only i.e. convert yolox model from pytorch to onnx and use it with opencv. Is there any way to improve the FPS?
Hey! You can try with nano, tiny, and small models. Performance will vary from device to device. It is recommended to test with the pre-trained ONNX models before custom training. Keep input images below 720p for optimal speed.
excellent
Top! Thanks.
How is this any different from centernet?
It is faster and more accurate than CenterNet. However there are new models for CenterNet like CenterNet2 and CenterNet++. We are yet to explore them in detail.