detect result

[zhanglaplace]

您好，这边测试了一个weights/checkpoint_VOC_efficientdet-d1_37.pth的模型，结果是这个样子的，有几个好像未检测到

[toandaominh1997]

hi @zhanglaplace, I just trained on VOC at 37 epoch, it really hasn't optimized yet, If you get better for detection, you can try this weights: weights at epoch 65 And this is results for this epoch You can try it.:v Anyway, I will public a better weight when complete training.

[zhanglaplace]

@toandaominh1997
Looking forward to your new model.

[deweyamer]

hello! can you provide your train loss figures if you complete the training? i have training 130 epoch on VOC, but it hasn't optimized yet thank you!

[xafarranxera]

Good morning,

I couldn't run the demo with the weights for D1 at 65 epoch, as checkpoint hasn't any "parser" inside

Has anybody got this solved?

hi @zhanglaplace, I just trained on VOC at 37 epoch, it really hasn't optimized yet, If you get better for detection, you can try this weights: weights at epoch 65 And this is results for this epoch You can try it.:v Anyway, I will public a better weight when complete training.

[inFreedom92]

@xafarranxera The code has been changed since the author replied, so you should retrain it your self. I think...

[xafarranxera]

Yes... I changed the code a bit to solve this, but I got a permission error for downloading the necessary efficientnet-d1 model. We're gonna wait for the d1 trained version from the author, luckily.

[inFreedom92]

@xafarranxera You could try this. #110

[xafarranxera]

@inFreedom92 Working now. Thank you so much.

[xafarranxera]

@inFreedom92 Regarding that, have you been able to get any good pretrained model for D1-D7 versions?

[inFreedom92]

@xafarranxera I'm currently training with my own dataset. I'm trying to detect cars and license plates. The model that inference below image is by using this repository's code about a month ago. But the author of this repository changed some code so I'm trying to catch up and make my model better. In my opinion, it is not good enough.