R-FCN/Faster-rcnn使用snapshot繼續訓練

r-fcn 和 faster-rcnn本身不能直接使用solverstate繼續訓練。
需要修改lib/fast_rcnn /train.py 和 tools/train_net.py
具體：
lib/fast_rcnn/train.py
修改init函數:

def __init__(self, solver_prototxt, roidb, output_dir,pretrained_model=None,##########add#########previous_state=None):"""Initialize the SolverWrapper."""self.output_dir = output_dirif (cfg.TRAIN.HAS_RPN and cfg.TRAIN.BBOX_REG andcfg.TRAIN.BBOX_NORMALIZE_TARGETS):# RPN can only use precomputed normalization because there are no# fixed statistics to compute a prioriassert cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTEDif cfg.TRAIN.BBOX_REG:print 'Computing bounding-box regression targets...'self.bbox_means, self.bbox_stds = \rdl_roidb.add_bbox_regression_targets(roidb)print 'done'self.solver = caffe.SGDSolver(solver_prototxt)if pretrained_model is not None:print ('Loading pretrained model ''weights from {:s}').format(pretrained_model)self.solver.net.copy_from(pretrained_model)##########add#########elif previous_state is not None:print ('Restoring State from {:s}').format(previous_state)self.solver.restore(previous_state)

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9
• 10
• 11
• 12
• 13
• 14
• 15
• 16
• 17
• 18
• 19
• 20
• 21
• 22
• 23
• 24
• 25
• 26
• 27
• 28
• 29

修改train_net函數：

def train_net(solver_prototxt, roidb, output_dir,pretrained_model=None, max_iters=40000,##########add#########previous_state=None):"""Train a Fast R-CNN network."""roidb = filter_roidb(roidb)sw = SolverWrapper(solver_prototxt, roidb, output_dir,##########add#########previous_state=previous_state)print 'Solving...'model_paths = sw.train_model(max_iters)print 'done solving'return model_paths

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9
• 10
• 11
• 12
• 13
• 14
• 15
• 16
• 17

tools/train_net.py:
修改prase_argsh函數，添加如下：

parser.add_argument('--snapshot', dest='previous_state',help='initialize with previous state',default=None, type=str)

• 1
• 2
• 3

修改main函數：

train_net(args.solver, roidb, output_dir,pretrained_model=args.pretrained_model,max_iters=args.max_iters,##########add#########previous_state=args.previous_state)

• 1
• 2
• 3
• 4
• 5
• 6

這樣就可以直接使用snapshot了。

總結

以上是生活随笔為你收集整理的R-FCN/Faster-rcnn使用snapshot继续训练的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。