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

updates

This commit is contained in:
Glenn Jocher 2018-11-22 13:52:22 +01:00
parent a46e500f9e
commit 809667404f
3 changed files with 37 additions and 35 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
test.py
View File

@ -16,7 +16,7 @@ parser.add_argument('-nms_thres', type=float, default=0.45, help='iou threshold
parser.add_argument('-n_cpu', type=int, default=0, help='number of cpu threads to use during batch generation') parser.add_argument('-n_cpu', type=int, default=0, help='number of cpu threads to use during batch generation')
parser.add_argument('-img_size', type=int, default=416, help='size of each image dimension') parser.add_argument('-img_size', type=int, default=416, help='size of each image dimension')
opt = parser.parse_args() opt = parser.parse_args()
print(opt) print(opt, end='\n\n')
cuda = torch.cuda.is_available() cuda = torch.cuda.is_available()
device = torch.device('cuda:0' if cuda else 'cpu') device = torch.device('cuda:0' if cuda else 'cpu')
@ -49,10 +49,8 @@ def main(opt):
# dataloader = torch.utils.data.DataLoader(dataset, batch_size=opt.batch_size, shuffle=False, num_workers=opt.n_cpu) # dataloader = torch.utils.data.DataLoader(dataset, batch_size=opt.batch_size, shuffle=False, num_workers=opt.n_cpu)
dataloader = load_images_and_labels(test_path, batch_size=opt.batch_size, img_size=opt.img_size) dataloader = load_images_and_labels(test_path, batch_size=opt.batch_size, img_size=opt.img_size)
print('Compute mAP...') print('%11s' * 5 % ('Image', 'Total', 'P', 'R', 'mAP'))
outputs, mAPs, mR, mP, TP, confidence, pred_class, target_class = [], [], [], [], [], [], [], []
mAP = 0
outputs, mAPs, TP, confidence, pred_class, target_class = [], [], [], [], [], []
AP_accum, AP_accum_count = np.zeros(nC), np.zeros(nC) AP_accum, AP_accum_count = np.zeros(nC), np.zeros(nC)
for batch_i, (imgs, targets) in enumerate(dataloader): for batch_i, (imgs, targets) in enumerate(dataloader):
imgs = imgs.to(device) imgs = imgs.to(device)
@ -107,22 +105,25 @@ def main(opt):
correct.append(0) correct.append(0)
# Compute Average Precision (AP) per class # Compute Average Precision (AP) per class
AP, AP_class = ap_per_class(tp=correct, conf=detections[:, 4], pred_cls=detections[:, 6], AP, AP_class, R, P = ap_per_class(tp=correct, conf=detections[:, 4], pred_cls=detections[:, 6],
target_cls=target_cls) target_cls=target_cls)
# Accumulate AP per class # Accumulate AP per class
AP_accum_count += np.bincount(AP_class, minlength=nC) AP_accum_count += np.bincount(AP_class, minlength=nC)
AP_accum += np.bincount(AP_class, minlength=nC, weights=AP) AP_accum += np.bincount(AP_class, minlength=nC, weights=AP)
# Compute mean AP for this image # Compute mean AP across all classes in this image, and append to image list
mAP = AP.mean() mAPs.append(AP.mean())
mR.append(R.mean())
mP.append(P.mean())
# Append image mAP to list # Means of all images
mAPs.append(mAP)
mean_mAP = np.mean(mAPs) mean_mAP = np.mean(mAPs)
mean_R = np.mean(mR)
mean_P = np.mean(mP)
# Print image mAP and running mean mAP # Print image mAP and running mean mAP
print('Image %d/%d AP: %.4f (%.4f)' % (len(mAPs), len(dataloader) * opt.batch_size, mAP, mean_mAP)) print(('%11s%11s' + '%11.3g' * 3) % (len(mAPs), len(dataloader) * opt.batch_size, mean_P, mean_R, mean_mAP))
# Print mAP per class # Print mAP per class
classes = load_classes(opt.class_path) # Extracts class labels from file classes = load_classes(opt.class_path) # Extracts class labels from file
@ -130,8 +131,8 @@ def main(opt):
print('%15s: %-.4f' % (c, AP_accum[i] / AP_accum_count[i])) print('%15s: %-.4f' % (c, AP_accum[i] / AP_accum_count[i]))
# Print mAP # Print mAP
print('Mean Average Precision: %.4f' % mean_mAP) print('%11s' * 5 % ('Image', 'Total', 'P', 'R', 'mAP'))
return mean_mAP return mean_mAP, mean_R, mean_P
if __name__ == '__main__': if __name__ == '__main__':

6
train.py
View File

@ -125,7 +125,7 @@ def main(opt):
g['lr'] = lr g['lr'] = lr
# Compute loss, compute gradient, update parameters # Compute loss, compute gradient, update parameters
loss = model(imgs.to(device), targets, requestPrecision=True) loss = model(imgs.to(device), targets, requestPrecision=False)
loss.backward() loss.backward()
# accumulated_batches = 1 # accumulate gradient for 4 batches before stepping optimizer # accumulated_batches = 1 # accumulate gradient for 4 batches before stepping optimizer
@ -183,11 +183,11 @@ def main(opt):
# Calculate mAP # Calculate mAP
import test import test
test.opt.weights_path = 'weights/latest.pt' test.opt.weights_path = 'weights/latest.pt'
mAP = test.main(test.opt) mAP, R, P = test.main(test.opt)
# Write epoch results # Write epoch results
with open('results.txt', 'a') as file: with open('results.txt', 'a') as file:
file.write(s + '%11.3g' % mAP + '\n') file.write(s + '%11.3g' * 3 % (mAP, P, R) + '\n')
# Save final model # Save final model
dt = time.time() - t0 dt = time.time() - t0

35
utils/utils.py
View File

@ -14,20 +14,20 @@ def load_classes(path):
""" """
Loads class labels at 'path' Loads class labels at 'path'
""" """
fp = open(path, "r") fp = open(path, 'r')
names = fp.read().split("\n")[:-1] names = fp.read().split('\n')[:-1]
return names return names
def model_info(model): # Plots a line-by-line description of a PyTorch model def model_info(model): # Plots a line-by-line description of a PyTorch model
nP = sum(x.numel() for x in model.parameters()) # number parameters n_p = sum(x.numel() for x in model.parameters()) # number parameters
nG = sum(x.numel() for x in model.parameters() if x.requires_grad) # number gradients n_g = sum(x.numel() for x in model.parameters() if x.requires_grad) # number gradients
print('\n%4s %70s %9s %12s %20s %12s %12s' % ('', 'name', 'gradient', 'parameters', 'shape', 'mu', 'sigma')) print('\n%4s %70s %9s %12s %20s %12s %12s' % ('', 'name', 'gradient', 'parameters', 'shape', 'mu', 'sigma'))
for i, (name, p) in enumerate(model.named_parameters()): for i, (name, p) in enumerate(model.named_parameters()):
name = name.replace('module_list.', '') name = name.replace('module_list.', '')
print('%4g %70s %9s %12g %20s %12g %12g' % ( print('%4g %70s %9s %12g %20s %12g %12g' % (
i, name, p.requires_grad, p.numel(), list(p.shape), p.mean(), p.std())) i, name, p.requires_grad, p.numel(), list(p.shape), p.mean(), p.std()))
print('\n%g layers, %g parameters, %g gradients' % (i + 1, nP, nG)) print('\nModel Summary: %g layers, %g parameters, %g gradients\n' % (i + 1, n_p, n_g))
def class_weights(): # frequency of each class in coco train2014 def class_weights(): # frequency of each class in coco train2014
@ -104,7 +104,7 @@ def ap_per_class(tp, conf, pred_cls, target_cls):
unique_classes = np.unique(np.concatenate((pred_cls, target_cls), 0)) unique_classes = np.unique(np.concatenate((pred_cls, target_cls), 0))
# Create Precision-Recall curve and compute AP for each class # Create Precision-Recall curve and compute AP for each class
ap = [] ap, p, r = [], [], []
for c in unique_classes: for c in unique_classes:
i = pred_cls == c i = pred_cls == c
n_gt = sum(target_cls == c) # Number of ground truth objects n_gt = sum(target_cls == c) # Number of ground truth objects
@ -112,25 +112,27 @@ def ap_per_class(tp, conf, pred_cls, target_cls):
if (n_p == 0) and (n_gt == 0): if (n_p == 0) and (n_gt == 0):
continue continue
elif (np == 0) and (n_gt > 0): elif (n_p == 0) or (n_gt == 0):
ap.append(0)
elif (n_p > 0) and (n_gt == 0):
ap.append(0) ap.append(0)
r.append(0)
p.append(0)
else: else:
# Accumulate FPs and TPs # Accumulate FPs and TPs
fpa = np.cumsum(1 - tp[i]) fpc = np.cumsum(1 - tp[i])
tpa = np.cumsum(tp[i]) tpc = np.cumsum(tp[i])
# Recall # Recall
recall = tpa / (n_gt + 1e-16) recall_curve = tpc / (n_gt + 1e-16)
r.append(tpc[-1] / (n_gt + 1e-16))
# Precision # Precision
precision = tpa / (tpa + fpa) precision_curve = tpc / (tpc + fpc)
p.append(tpc[-1] / (tpc[-1] + fpc[-1]))
# AP from recall-precision curve # AP from recall-precision curve
ap.append(compute_ap(recall, precision)) ap.append(compute_ap(recall_curve, precision_curve))
return np.array(ap), unique_classes.astype('int32') return np.array(ap), unique_classes.astype('int32'), np.array(r), np.array(p)
def compute_ap(recall, precision): def compute_ap(recall, precision):
@ -431,7 +433,7 @@ def coco_class_count(path='/Users/glennjocher/downloads/DATA/coco/labels/train20
def plot_results(): def plot_results():
# Plot YOLO training results file "results.txt" # Plot YOLO training results file 'results.txt'
import numpy as np import numpy as np
import matplotlib.pyplot as plt import matplotlib.pyplot as plt
plt.figure(figsize=(16, 8)) plt.figure(figsize=(16, 8))
@ -445,4 +447,3 @@ def plot_results():
plt.title(s[i]) plt.title(s[i])
if i == 0: if i == 0:
plt.legend() plt.legend()