leon

A problem about loss and lr value?

1

In your code, i see in the loss.py the final loss value is the value after excluding twice the batchsize . And in your train.py，final lr multiply the batchsize. so...

🔥🔥🔥some questions?

4

1、why sceneflow dataset need to pad in the test, but kitti2015 don't need? 2、code about computing 3-px error ` true_disp = disp_true` their memory address are same, so `correct =...

about time in Table 4. The KITTI 2015 leaderboard presented?

1

in the Table 4, PSMNet's time is 0.41......so, i want to know the your environment (use gpu or cpu?),and this time is got in the "Test_img.py" or in the "submission.py"(in...

some problems?

1

1、why make the png to npy? 2、when i make the json‘s gpu_id=[0,1],it only support the batchsize=2?when i make the batchsize=4, out of memory?but the gpu only use 1G/10G？i don't know...

question about mAP ??????

2

on my test : 1、VOC=2007test、yolov3 、 conf threshold =0.005 ,mAP is 0.74 2、VOC=2007test、yolov3、conf threshold=0.01,mAP is 0.52. 【mAP's function is "faster rcnn ---voc_eval.py"】 but in https://github.com/Stinky-Tofu/Stronger-yolo/blob/master/v2/README.md, conf_threshold=0.01,but mAP is 0.733(baseline, no...

how to solve this problem?

1