YOLOX
YOLOX copied to clipboard
Megvii-BaseDetection
How can i know the all information of YOLOX evalution, such as TP,FP,FN,Prcision,Recall,etc?
I have changed parameters in coco_evaluator.py ,voc_eval.py and voc_evaluator.py, but there's still no something i wanna get...
like follwing pic I assigned a .txt to store the information, but it failed...I really have noooo idea how to resolve my issue.
I mean mAP is not enough to my experiment, i wanna get tp,fp,fn, precision,recall and AP of each classes.
Evaluation in yolox coco only contains Precision, recall and mAP. You need to write your own code to get the value of TP,FP,FN.
@FateScript @Joker316701882 How can i print the precision and recall values? please help needed
@yolor2 functions named per_class_AR_table and per_class_AP_table in yolox.evaluator might help you.
@FateScript So in voc_eval.py I get my recall value in np.ndarray type. And to get the recall/precision values I just need to mean the array?
I just modify two files. One of that name coco_evaluator.py is located at yolox/yolox/evaluators, and the other is located at yolox/yolox/fast_coco_eval_api.py
You can try it.
Hello, I have the same needs as yours, I need information such as TP, FP, FN, TN, so can you tell me how to modify the file, thank you
TypeError: evaluate() got an unexpected keyword argument 'return_outputs How to solve this problem thanks
I just modify two files.
One of that name coco_evaluator.py is located at yolox/yolox/evaluators, and the other is located at yolox/yolox/fast_coco_eval_api.py
You can try it.
I think I wrote the wrong folder,sorry! The correct path is that:
1⃣️ yolox/evaluators/coco_evaluator.py 2⃣️ yolox/layers/fast_coco_eval.api.py
yolox.zip I just modify two files. One of that name coco_evaluator.py is located at yolox/yolox/evaluators, and the other is located at yolox/yolox/fast_coco_eval_api.py You can try it.
I think I wrote the wrong folder,sorry! The correct path is that:
1⃣️ yolox/evaluators/coco_evaluator.py 2⃣️ yolox/layers/fast_coco_eval.api.py
ok，I found this problem and solved it, but a new problem arose AttributeError: 'bool' object has no attribute 'seek'. You can only torch.load from a file that is seekable. Please pre-load the data into a buffer like io.BytesIO and try to load from it instead.
约洛克斯.zip 我只是修改了两个文件。 其中一个名称coco_evaluator.py位于yolox/yolox/evaluators，另一个位于yolox/yolox/fast_coco_eval_api.py 你可以试试。
我想我写错了文件夹，对不起！正确的路径是：
1⃣️ yolox/评估员/coco_evaluator.py 2⃣️ yolox/layers/fast_coco_eval.api.py
By the way, I actually run it on Google's colab
https://www.796t.com/content/1549239487.html
I think this article can figure out your problem!
https://www.796t.com/content/1549239487.html
I think this article can figure out your problem!
我想再次詢問您的test.json文件是什麼
它是個存放訓練集的一個檔案，如圖所示
你可以參考此文了解資料集的常見格式 https://github.com/huchi00057/Dataset
它是個存放訓練集的一個檔案，如圖所示
你可以參考此文了解資料集的常見格式 https://github.com/huchi00057/Dataset
我這邊還是報錯，可否把數據集給您，幫我訓練一下，有償請您説明。 實在是需要TP，FP等結果
约洛克斯.zip 我只是修改了两个文件。 其中一个名称coco_evaluator.py位于yolox/yolox/evaluators，另一个位于yolox/yolox/fast_coco_eval_api.py 你可以试试。
我想我写错了文件夹，对不起！正确的路径是： 1⃣️ yolox/评估员/coco_evaluator.py 2⃣️ yolox/layers/fast_coco_eval.api.py
By the way, I actually run it on Google's colab
除了这两个文件，其他的文件没有更改的吗？
Sorry for interpretin your talk. When I tried your coco evaluator, the number of results of FP is unbelivable. ` per class tp @0.50:
| class | # | class | # | class | # |
|---|---|---|---|---|---|
| Can | 280.000 | Bottle | 926.000 | Box | 172.000 |
| Cigarette | 530.000 | ||||
| per class fp @0.50: | |||||
| class | # | class | # | class | # |
| :---------- | :--------- | :-------- | :--------- | :-------- | :-------- |
| Can | 1408.000 | Bottle | 2618.000 | Box | 860.000 |
| Cigarette | 4772.000 | ||||
| per class fn @0.50: | |||||
| class | # | class | # | class | # |
| :---------- | :------- | :-------- | :------- | :-------- | :------ |
| Can | 7.000 | Bottle | 15.000 | Box | 7.000 |
| Cigarette | 15.000 |
` So, Would you tell me how to calc fp in your code? I'm trying to figure out your code now. Thank you in advance.
