Bienqq
/
car-detection-bayes
2
1
Fork 0
Code Issues Pull Requests Releases Wiki Activity

updates

This commit is contained in:
Glenn Jocher 2019-04-12 14:00:16 +02:00
parent d5db50df8e
commit bce3dd03e8
3 changed files with 10 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
models.py
View File

@ -33,7 +33,7 @@ def create_modules(module_defs):
if bn: if bn:
modules.add_module('batch_norm_%d' % i, nn.BatchNorm2d(filters)) modules.add_module('batch_norm_%d' % i, nn.BatchNorm2d(filters))
if module_def['activation'] == 'leaky': if module_def['activation'] == 'leaky':
modules.add_module('leaky_%d' % i, nn.LeakyReLU(0.1)) modules.add_module('leaky_%d' % i, nn.LeakyReLU(0.1, inplace=True))
elif module_def['type'] == 'maxpool': elif module_def['type'] == 'maxpool':
kernel_size = int(module_def['size']) kernel_size = int(module_def['size'])

3
train.py
View File

@ -49,7 +49,6 @@ def train(
start_epoch = 0 start_epoch = 0
best_loss = float('inf') best_loss = float('inf')
nf = int(model.module_defs[model.yolo_layers[0] - 1]['filters']) # yolo layer size (i.e. 255) nf = int(model.module_defs[model.yolo_layers[0] - 1]['filters']) # yolo layer size (i.e. 255)
if resume: # Load previously saved model if resume: # Load previously saved model
if transfer: # Transfer learning if transfer: # Transfer learning
chkpt = torch.load(weights + 'yolov3.pt', map_location=device) chkpt = torch.load(weights + 'yolov3.pt', map_location=device)
@ -74,7 +73,7 @@ def train(
else: else:
cutoff = load_darknet_weights(model, weights + 'darknet53.conv.74') cutoff = load_darknet_weights(model, weights + 'darknet53.conv.74')
# Set scheduler (reduce lr at epochs 218, 245, i.e. batches 400k, 450k) # Scheduler (reduce lr at epochs 218, 245, i.e. batches 400k, 450k)
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=[218, 245], gamma=0.1, scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=[218, 245], gamma=0.1,
last_epoch=start_epoch - 1) last_epoch=start_epoch - 1)

9
utils/utils.py
View File

@ -326,7 +326,6 @@ def build_targets(model, targets):
return txy, twh, tcls, indices return txy, twh, tcls, indices
# @profile
def non_max_suppression(prediction, conf_thres=0.5, nms_thres=0.5): def non_max_suppression(prediction, conf_thres=0.5, nms_thres=0.5):
""" """
Removes detections with lower object confidence score than 'conf_thres' Removes detections with lower object confidence score than 'conf_thres'
@ -383,6 +382,11 @@ def non_max_suppression(prediction, conf_thres=0.5, nms_thres=0.5):
dc = pred[pred[:, -1] == c] # select class c dc = pred[pred[:, -1] == c] # select class c
dc = dc[:min(len(dc), 100)] # limit to first 100 boxes: https://github.com/ultralytics/yolov3/issues/117 dc = dc[:min(len(dc), 100)] # limit to first 100 boxes: https://github.com/ultralytics/yolov3/issues/117
# No NMS required if only 1 prediction
if len(dc) == 1:
det_max.append(dc)
continue
# Non-maximum suppression # Non-maximum suppression
if nms_style == 'OR': # default if nms_style == 'OR': # default
# METHOD1 # METHOD1
@ -410,6 +414,9 @@ def non_max_suppression(prediction, conf_thres=0.5, nms_thres=0.5):
elif nms_style == 'MERGE': # weighted mixture box elif nms_style == 'MERGE': # weighted mixture box
while len(dc): while len(dc):
if len(dc) == 1:
det_max.append(dc)
break
i = bbox_iou(dc[0], dc) > nms_thres # iou with other boxes i = bbox_iou(dc[0], dc) > nms_thres # iou with other boxes
weights = dc[i, 4:5] weights = dc[i, 4:5]
dc[0, :4] = (weights * dc[i, :4]).sum(0) / weights.sum() dc[0, :4] = (weights * dc[i, :4]).sum(0) / weights.sum()