thanks for your great work!But i meet a problem and i want to get your help!When i change the backbone from densenet121.features to vovnet57 or resnet50.My testing results are really bad which almost 0.And if i don't change the lr from 0.004 to smaller,it will show as below.
iter: 250, acc (bg: 0.96, fg: 0.00, iou: 0.31), loss (bbox_3d: 10.6909, cls: 0.2957, iou: 1.1781), misc (ry: 4.16, z: 7.97), dt: 0.86, eta: 11.9h
iter: 500, acc (bg: 1.00), loss (cls: 0.0065), dt: 0.80, eta: 11.1h
iter: 750, acc (bg: 1.00, fg: 0.00, iou: 0.25), loss (bbox_3d: 37.9193, cls: 0.0167, iou: 1.4007), misc (ry: 9.57, z: 24.60), dt: 0.84, eta: 11.5h
Some information losses.
i want to know if you did some ecxperiments to change the backbone.
Thank you
thanks for your great work!But i meet a problem and i want to get your help!When i change the backbone from densenet121.features to vovnet57 or resnet50.My testing results are really bad which almost 0.And if i don't change the lr from 0.004 to smaller,it will show as below.
iter: 250, acc (bg: 0.96, fg: 0.00, iou: 0.31), loss (bbox_3d: 10.6909, cls: 0.2957, iou: 1.1781), misc (ry: 4.16, z: 7.97), dt: 0.86, eta: 11.9h
iter: 500, acc (bg: 1.00), loss (cls: 0.0065), dt: 0.80, eta: 11.1h
iter: 750, acc (bg: 1.00, fg: 0.00, iou: 0.25), loss (bbox_3d: 37.9193, cls: 0.0167, iou: 1.4007), misc (ry: 9.57, z: 24.60), dt: 0.84, eta: 11.5h
Some information losses.
i want to know if you did some ecxperiments to change the backbone.
Thank you