Pre-trained caffe model for FCN score evaluation

[Noah-Jrh]

I tried your evaluation code. I change the parameter --result_dir in ./scripts/eval_cityscapes/evaluate.py to the original val set of Cityscapes. I have resized all the original Cityscapes
images(*_gtFine_labelIds.png and *_leftImg8bit.png) to 256x256, but the segmentation results become so bad as follows：
Mean pixel accuracy: 0.394909
Mean class accuracy: 0.057150
Mean class IoU: 0.025087
************ Per class numbers below ************
road : acc = 0.999244, iou = 0.390848
sidewalk : acc = 0.001727, iou = 0.001507
building : acc = 0.062116, iou = 0.061576
wall : acc = 0.000011, iou = 0.000011
fence : acc = 0.000008, iou = 0.000008
pole : acc = 0.000036, iou = 0.000036
traffic light : acc = 0.000000, iou = 0.000000
traffic sign : acc = 0.000000, iou = 0.000000
vegetation : acc = 0.000003, iou = 0.000003
terrain : acc = 0.000000, iou = 0.000000
sky : acc = 0.022368, iou = 0.022336
person : acc = 0.000000, iou = 0.000000
rider : acc = 0.000000, iou = 0.000000
car : acc = 0.000329, iou = 0.000322
truck : acc = 0.000000, iou = 0.000000
bus : acc = 0.000000, iou = 0.000000
train : acc = 0.000000, iou = 0.000000
motorcycle : acc = 0.000000, iou = 0.000000
bicycle : acc = 0.000000, iou = 0.000000
Is there anything wrong with this fcn-8s-cityscapes.caffemodel ?

[junyanz]

Please see these two posts 1 and 2
@tinghuiz