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

updates

This commit is contained in:
Glenn Jocher 2019-10-05 15:28:02 +02:00
parent 563dad3b53
commit b6d9a742ec
2 changed files with 8 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
train.py
View File

@ -20,6 +20,10 @@ last = wdir + 'last.pt'
best = wdir + 'best.pt' best = wdir + 'best.pt'
results_file = 'results.txt' results_file = 'results.txt'
# last = '../coco/last_sgd273_default_ltrb25x4.pt'
# best = '../coco/best_sgd273_default_ltrb25x4.pt'
# results_file = '../coco/results_sgd273_default_ltrb25x4.txt'
# Hyperparameters (j-series, 50.5 mAP yolov3-320) evolved by @ktian08 https://github.com/ultralytics/yolov3/issues/310 # Hyperparameters (j-series, 50.5 mAP yolov3-320) evolved by @ktian08 https://github.com/ultralytics/yolov3/issues/310
hyp = {'giou': 1.582, # giou loss gain hyp = {'giou': 1.582, # giou loss gain
'cls': 27.76, # cls loss gain (CE=~1.0, uCE=~20) 'cls': 27.76, # cls loss gain (CE=~1.0, uCE=~20)
@ -386,8 +390,8 @@ if __name__ == '__main__':
parser.add_argument('--epochs', type=int, default=273) # 500200 batches at bs 16, 117263 images = 273 epochs parser.add_argument('--epochs', type=int, default=273) # 500200 batches at bs 16, 117263 images = 273 epochs
parser.add_argument('--batch-size', type=int, default=32) # effective bs = batch_size * accumulate = 16 * 4 = 64 parser.add_argument('--batch-size', type=int, default=32) # effective bs = batch_size * accumulate = 16 * 4 = 64
parser.add_argument('--accumulate', type=int, default=2, help='batches to accumulate before optimizing') parser.add_argument('--accumulate', type=int, default=2, help='batches to accumulate before optimizing')
parser.add_argument('--cfg', type=str, default='cfg/yolov3-spp.cfg', help='cfg file path') parser.add_argument('--cfg', type=str, default='cfg/yolov3-spp-2overlap.cfg', help='cfg file path')
parser.add_argument('--data', type=str, default='data/coco.data', help='*.data file path') parser.add_argument('--data', type=str, default='data/coco_64img.data', help='*.data file path')
parser.add_argument('--multi-scale', action='store_true', help='adjust (67% - 150%) img_size every 10 batches') parser.add_argument('--multi-scale', action='store_true', help='adjust (67% - 150%) img_size every 10 batches')
parser.add_argument('--img-size', type=int, default=416, help='inference size (pixels)') parser.add_argument('--img-size', type=int, default=416, help='inference size (pixels)')
parser.add_argument('--rect', action='store_true', help='rectangular training') parser.add_argument('--rect', action='store_true', help='rectangular training')

4
utils/datasets.py
View File

@ -644,7 +644,7 @@ def cutout(image, labels):
return inter_area / box2_area return inter_area / box2_area
# create random masks # create random masks
scales = [0.5] * 1 # + [0.25] * 4 + [0.125] * 16 # image size fraction scales = [0.5] * 1 # + [0.25] * 4 + [0.125] * 16 + [0.0625] * 64 + [0.03125] * 256 # image size fraction
for s in scales: for s in scales:
mask_h = random.randint(1, int(h * s)) mask_h = random.randint(1, int(h * s))
mask_w = random.randint(1, int(w * s)) mask_w = random.randint(1, int(w * s))
@ -660,7 +660,7 @@ def cutout(image, labels):
image[ymin:ymax, xmin:xmax] = mask_color image[ymin:ymax, xmin:xmax] = mask_color
# return unobscured labels # return unobscured labels
if len(labels): if len(labels) and s > 0.03:
box = np.array([xmin, ymin, xmax, ymax], dtype=np.float32) box = np.array([xmin, ymin, xmax, ymax], dtype=np.float32)
ioa = bbox_ioa(box, labels[:, 1:5]) # intersection over area ioa = bbox_ioa(box, labels[:, 1:5]) # intersection over area
labels = labels[ioa < 0.90] # remove >90% obscured labels labels = labels[ioa < 0.90] # remove >90% obscured labels