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

training updates

This commit is contained in:
Glenn Jocher 2020-04-08 21:01:58 -07:00
parent c7f5b6cc21
commit 4120ac3aa6
3 changed files with 8 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
train.py
View File

@ -209,7 +209,7 @@ def train():
# Start training # Start training
nb = len(dataloader) # number of batches nb = len(dataloader) # number of batches
n_burn = max(3 * nb, 300) # burn-in iterations, max(3 epochs, 300 iterations) n_burn = max(3 * nb, 500) # burn-in iterations, max(3 epochs, 300 iterations)
maps = np.zeros(nc) # mAP per class maps = np.zeros(nc) # mAP per class
# torch.autograd.set_detect_anomaly(True) # torch.autograd.set_detect_anomaly(True)
results = (0, 0, 0, 0, 0, 0, 0) # 'P', 'R', 'mAP', 'F1', 'val GIoU', 'val Objectness', 'val Classification' results = (0, 0, 0, 0, 0, 0, 0) # 'P', 'R', 'mAP', 'F1', 'val GIoU', 'val Objectness', 'val Classification'
@ -234,8 +234,8 @@ def train():
targets = targets.to(device) targets = targets.to(device)
# Burn-in # Burn-in
if ni <= n_burn: if ni <= n_burn * 2:
model.gr = np.interp(ni, [0, n_burn], [0.0, 1.0]) # giou loss ratio (obj_loss = 1.0 or giou) model.gr = np.interp(ni, [0, n_burn * 2], [0.0, 1.0]) # giou loss ratio (obj_loss = 1.0 or giou)
if ni == n_burn: # burnin complete if ni == n_burn: # burnin complete
print_model_biases(model) print_model_biases(model)

2
utils/torch_utils.py
View File

@ -149,7 +149,7 @@ class ModelEMA:
self.ema = deepcopy(model) self.ema = deepcopy(model)
self.ema.eval() self.ema.eval()
self.updates = 0 # number of EMA updates self.updates = 0 # number of EMA updates
self.decay = lambda x: decay * (1 - math.exp(-x / 1000)) # decay exponential ramp (to help early epochs) self.decay = lambda x: decay * (1 - math.exp(-x / 2000)) # decay exponential ramp (to help early epochs)
self.device = device # perform ema on different device from model if set self.device = device # perform ema on different device from model if set
if device: if device:
self.ema.to(device=device) self.ema.to(device=device)

8
utils/utils.py
View File

@ -688,11 +688,11 @@ def coco_single_class_labels(path='../coco/labels/train2014/', label_class=43):
shutil.copyfile(src=img_file, dst='new/images/' + Path(file).name.replace('txt', 'jpg')) # copy images shutil.copyfile(src=img_file, dst='new/images/' + Path(file).name.replace('txt', 'jpg')) # copy images
def kmean_anchors(path='../coco/train2017.txt', n=9, img_size=(608, 608)): def kmean_anchors(path='../coco/train2017.txt', n=12, img_size=(320, 1024)):
# from utils.utils import *; _ = kmean_anchors() # from utils.utils import *; _ = kmean_anchors()
# Produces a list of target kmeans suitable for use in *.cfg files # Creaters kmeans anchors for use in *.cfg files
from utils.datasets import LoadImagesAndLabels from utils.datasets import LoadImagesAndLabels
thr = 0.20 # IoU threshold thr = 0.225 # IoU threshold
def print_results(k): def print_results(k):
k = k[np.argsort(k.prod(1))] # sort small to large k = k[np.argsort(k.prod(1))] # sort small to large
@ -709,7 +709,7 @@ def kmean_anchors(path='../coco/train2017.txt', n=9, img_size=(608, 608)):
def fitness(k): # mutation fitness def fitness(k): # mutation fitness
iou = wh_iou(wh, torch.Tensor(k)) # iou iou = wh_iou(wh, torch.Tensor(k)) # iou
max_iou = iou.max(1)[0] max_iou = iou.max(1)[0]
return max_iou.mean() # product return (max_iou * (max_iou > thr).float()).mean() # product
# Get label wh # Get label wh
wh = [] wh = []