params.json

{
  "name": "TffRCNN resnet50",
  "tagline": "Faster rcnn based on tensorflow and resnet50",
  "body": "# TFFRCN_Resnet50\r\n\r\nThis is the faster rcnn based on Tensorflow and resnet50\r\n\r\n\r\n  - fork from the TFFRCN website is https://github.com/CharlesShang/TFFRCNN\r\n  - just change the python file factory.py ， __ init __.py ，networks.py in /lib/networks \r\n  - add resnet50 networks \r\n  - resnet101 hasn't been tested\r\n\r\nYou can find how to use it from the above website\r\n\r\nthe model file is convert from the caffemodel, you can download from the baiduyun  \r\nlink: http://pan.baidu.com/s/1eSuUO1s pwd: 24cf  \r\nthe fine-tune model link: http://pan.baidu.com/s/1nuUYfMh pwd：5bve  \r\nresnet101 imagenet file link http://pan.baidu.com/s/1i5odxNv pwd:d8y4  \r\n\r\nUSing voc07_trainval to train the Resnet50 and test on the voc07_test  \r\nThe result is :\r\nMean AP = 0.7124\r\n\r\nAP for aeroplane = 0.7801  \r\nAP for bicycle = 0.7931  \r\nAP for bird = 0.6836  \r\nAP for boat = 0.5750  \r\nAP for bottle = 0.4892  \r\nAP for bus = 0.8322  \r\nAP for car = 0.8412  \r\nAP for cat = 0.8380  \r\nAP for chair = 0.5186  \r\nAP for cow = 0.7550  \r\nAP for diningtable = 0.6369  \r\nAP for dog = 0.7861  \r\nAP for horse = 0.7966  \r\nAP for motorbike = 0.7692  \r\nAP for person = 0.7765  \r\nAP for pottedplant = 0.4454  \r\nAP for sheep = 0.7131  \r\nAP for sofa = 0.7074  \r\nAP for train = 0.8150  \r\nAP for tvmonitor = 0.6953  \r\n\r\n\r\n",
  "note": "Don't delete this file! It's used internally to help with page regeneration."
}