This commit is contained in:
Glenn Jocher 2020-02-02 09:19:44 -08:00
parent 8b18beb3db
commit 785bfec286
1 changed files with 19 additions and 13 deletions

View File

@ -751,22 +751,22 @@ def kmean_anchors(path='../coco/train2017.txt', n=9, img_size=(320, 640)):
from utils.datasets import LoadImagesAndLabels from utils.datasets import LoadImagesAndLabels
thr = 0.20 # IoU threshold thr = 0.20 # IoU threshold
def print_results(thr, wh, k): def print_results(wh, k):
k = k[np.argsort(k.prod(1))] # sort small to large k = k[np.argsort(k.prod(1))] # sort small to large
iou = wh_iou(torch.Tensor(wh), torch.Tensor(k)) iou = wh_iou(torch.Tensor(wh), torch.Tensor(k))
max_iou, min_iou = iou.max(1)[0], iou.min(1)[0] max_iou = iou.max(1)[0]
bpr, aat = (max_iou > thr).float().mean(), (iou > thr).float().mean() * n # best possible recall, anch > thr bpr, aat = (max_iou > thr).float().mean(), (iou > thr).float().mean() * n # best possible recall, anch > thr
print('%.2f iou_thr: %.3f best possible recall, %.2f anchors > thr' % (thr, bpr, aat)) print('%.2f iou_thr: %.3f best possible recall, %.2f anchors > thr' % (thr, bpr, aat))
print('kmeans anchors (n=%g, img_size=%s, IoU=%.3f/%.3f/%.3f-min/mean/best): ' % print('n=%g, img_size=%s, IoU_all=%.3f/%.3f-mean/best, IoU>thr=%.3f-mean: ' %
(n, img_size, min_iou.mean(), iou.mean(), max_iou.mean()), end='') (n, img_size, iou.mean(), max_iou.mean(), iou[iou > thr].mean()), end='')
for i, x in enumerate(k): for i, x in enumerate(k):
print('%i,%i' % (round(x[0]), round(x[1])), end=', ' if i < len(k) - 1 else '\n') # use in *.cfg print('%i,%i' % (round(x[0]), round(x[1])), end=', ' if i < len(k) - 1 else '\n') # use in *.cfg
return k return k
def fitness(thr, wh, k): # mutation fitness def fitness(wh, k): # mutation fitness
iou = wh_iou(wh, torch.Tensor(k)).max(1)[0] # max iou iou = wh_iou(wh, torch.Tensor(k)) # iou
bpr = (iou > thr).float().mean() # best possible recall max_iou = iou.max(1)[0]
return iou.mean() * bpr # product return max_iou.mean() # product
# Get label wh # Get label wh
wh = [] wh = []
@ -776,6 +776,7 @@ def kmean_anchors(path='../coco/train2017.txt', n=9, img_size=(320, 640)):
wh.append(l[:, 3:5] * (s / s.max())) # image normalized to letterbox normalized wh wh.append(l[:, 3:5] * (s / s.max())) # image normalized to letterbox normalized wh
wh = np.concatenate(wh, 0).repeat(nr, axis=0) # augment 10x wh = np.concatenate(wh, 0).repeat(nr, axis=0) # augment 10x
wh *= np.random.uniform(img_size[0], img_size[1], size=(wh.shape[0], 1)) # normalized to pixels (multi-scale) wh *= np.random.uniform(img_size[0], img_size[1], size=(wh.shape[0], 1)) # normalized to pixels (multi-scale)
wh = wh[(wh > 2.0).all(1)] # remove below threshold boxes (< 2 pixels wh)
# Darknet yolov3.cfg anchors # Darknet yolov3.cfg anchors
use_darknet = False use_darknet = False
@ -788,7 +789,7 @@ def kmean_anchors(path='../coco/train2017.txt', n=9, img_size=(320, 640)):
s = wh.std(0) # sigmas for whitening s = wh.std(0) # sigmas for whitening
k, dist = kmeans(wh / s, n, iter=30) # points, mean distance k, dist = kmeans(wh / s, n, iter=30) # points, mean distance
k *= s k *= s
k = print_results(thr, wh, k) k = print_results(wh, k)
# # Plot # # Plot
# k, d = [None] * 20, [None] * 20 # k, d = [None] * 20, [None] * 20
@ -797,21 +798,26 @@ def kmean_anchors(path='../coco/train2017.txt', n=9, img_size=(320, 640)):
# fig, ax = plt.subplots(1, 2, figsize=(14, 7)) # fig, ax = plt.subplots(1, 2, figsize=(14, 7))
# ax = ax.ravel() # ax = ax.ravel()
# ax[0].plot(np.arange(1, 21), np.array(d) ** 2, marker='.') # ax[0].plot(np.arange(1, 21), np.array(d) ** 2, marker='.')
# fig, ax = plt.subplots(1, 2, figsize=(14, 7)) # plot wh
# ax[0].hist(wh[wh[:, 0]<100, 0],400)
# ax[1].hist(wh[wh[:, 1]<100, 1],400)
# fig.tight_layout()
# fig.savefig('wh.png', dpi=200)
# Evolve # Evolve
npr = np.random npr = np.random
wh = torch.Tensor(wh) wh = torch.Tensor(wh)
f, sh, ng, mp, s = fitness(thr, wh, k), k.shape, 1000, 0.9, 0.1 # fitness, generations, mutation probability, sigma f, sh, ng, mp, s = fitness(wh, k), k.shape, 1000, 0.9, 0.1 # fitness, generations, mutation prob, sigma
for _ in tqdm(range(ng), desc='Evolving anchors'): for _ in tqdm(range(ng), desc='Evolving anchors'):
v = np.ones(sh) v = np.ones(sh)
while (v == 1).all(): # mutate until a change occurs (prevent duplicates) while (v == 1).all(): # mutate until a change occurs (prevent duplicates)
v = ((npr.random(sh) < mp) * npr.random() * npr.randn(*sh) * s + 1).clip(0.3, 3.0) # 98.6, 61.6 v = ((npr.random(sh) < mp) * npr.random() * npr.randn(*sh) * s + 1).clip(0.3, 3.0) # 98.6, 61.6
kg = (k.copy() * v).clip(min=2.0) kg = (k.copy() * v).clip(min=2.0)
fg = fitness(thr, wh, kg) fg = fitness(wh, kg)
if fg > f: if fg > f:
f, k = fg, kg.copy() f, k = fg, kg.copy()
print_results(thr, wh, k) print_results(wh, k)
k = print_results(thr, wh, k) k = print_results(wh, k)
return k return k