KARTIK SIRWANI

hey @zoezhou1999 with my personal experience I have seen just normal l1 loss on alpha should be good enough to get initial decent results..also in the paper you can see...

where is the FG extension code @zoezhou1999. When I open the link in Readme it gives 404.

Also do you take average of pixels for computing loss or is it just sum of differences of alpha for each pixel ?

Yes I faced it too..when I tried with tiny yolo model and custom anchors

Followed the same steps suggested by @veralauee . While category predictions seem to change with input image, attribute predictions are same regardless of input images

I used the source code for version 1.2 for mmdetection and now I am getting the error No module named 'mmcv.cnn.weight_init' . My mmcv version is :- mmcv==1.0.4 Other versions:...

Anyone else facing the same issue can update their mmcv version to mmcv==0.5.1. This fixed the issue for me. Also dont forget to use mmdetection v 1.2.0. Cheers !!!

Same issue for me . I am hitting the endpoint /api/{routename}/path. It says api not found

Also tried to run on coco dataset Got a result like this { "image_id":139, "category_id":62, "bbox":[ 292.0556945800781, 216.36961364746094, 61.314727783203125, 103.17518615722656 ], "score":0.9969334602355957, "segmentation":{ "size":[ 426, 640 ], "counts":"TUj3l0h;e0C=@a0NO3:F6JM3M3M3O1N1UOZEQOi:n0k0N1O1O10000001O0010O010O3M5K4L1018Ga0_O=B0NG;G9ZOk0K0N3N101N101O0000001O00SD@P;?nDMi:3UE1i:OWEAJLP;c0RE]Od;e0XD\\Oj;R100001EnCYOW