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

updates

This commit is contained in:
Glenn Jocher 2018-10-10 16:16:17 +02:00
parent d748bedb1d
commit f79e7ffa76
3 changed files with 30 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
models.py
View File

@ -93,7 +93,7 @@ class YOLOLayer(nn.Module):
stride = 8 stride = 8
# Build anchor grids # Build anchor grids
nG = int(self.img_dim / stride) nG = int(self.img_dim / stride) # number grid points
self.grid_x = torch.arange(nG).repeat(nG, 1).view([1, 1, nG, nG]).float() self.grid_x = torch.arange(nG).repeat(nG, 1).view([1, 1, nG, nG]).float()
self.grid_y = torch.arange(nG).repeat(nG, 1).t().view([1, 1, nG, nG]).float() self.grid_y = torch.arange(nG).repeat(nG, 1).t().view([1, 1, nG, nG]).float()
self.scaled_anchors = torch.FloatTensor([(a_w / stride, a_h / stride) for a_w, a_h in anchors]) self.scaled_anchors = torch.FloatTensor([(a_w / stride, a_h / stride) for a_w, a_h in anchors])
@ -139,7 +139,7 @@ class YOLOLayer(nn.Module):
if targets is not None: if targets is not None:
MSELoss = nn.MSELoss(size_average=True) MSELoss = nn.MSELoss(size_average=True)
BCEWithLogitsLoss = nn.BCEWithLogitsLoss(size_average=True) BCEWithLogitsLoss = nn.BCEWithLogitsLoss(size_average=True)
# CrossEntropyLoss = nn.CrossEntropyLoss() CrossEntropyLoss = nn.CrossEntropyLoss()
if requestPrecision: if requestPrecision:
gx = self.grid_x[:, :, :nG, :nG] gx = self.grid_x[:, :, :nG, :nG]
@ -156,7 +156,7 @@ class YOLOLayer(nn.Module):
if x.is_cuda: if x.is_cuda:
tx, ty, tw, th, mask, tcls = tx.cuda(), ty.cuda(), tw.cuda(), th.cuda(), mask.cuda(), tcls.cuda() tx, ty, tw, th, mask, tcls = tx.cuda(), ty.cuda(), tw.cuda(), th.cuda(), mask.cuda(), tcls.cuda()
# Mask outputs to ignore non-existing objects (but keep confidence predictions) # Compute losses
nT = sum([len(x) for x in targets]) # number of targets nT = sum([len(x) for x in targets]) # number of targets
nM = mask.sum().float() # number of anchors (assigned to targets) nM = mask.sum().float() # number of anchors (assigned to targets)
nB = len(targets) # batch size nB = len(targets) # batch size

2
train.py
View File

@ -85,7 +85,7 @@ def main(opt):
# Set scheduler # Set scheduler
# scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=[54, 61], gamma=0.1) # scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=[54, 61], gamma=0.1)
modelinfo(model) model_info(model)
t0, t1 = time.time(), time.time() t0, t1 = time.time(), time.time()
mean_recall, mean_precision = 0, 0 mean_recall, mean_precision = 0, 0
print('%10s' * 16 % ( print('%10s' * 16 % (

45
utils/utils.py
View File

@ -19,24 +19,26 @@ def load_classes(path):
return names return names
def modelinfo(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
nparams = sum(x.numel() for x in model.parameters()) nP = sum(x.numel() for x in model.parameters()) # number parameters
ngradients = sum(x.numel() for x in model.parameters() if x.requires_grad) nG = 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, nparams, ngradients)) print('\n%g layers, %g parameters, %g gradients' % (i + 1, nP, nG))
def xview_class_weights(indices): # weights of each class in the training set, normalized to mu = 1 def class_weights(): # frequency of each class in coco train2014
weights = 1 / torch.FloatTensor( weights = 1 / torch.FloatTensor(
[74, 364, 713, 71, 2925, 209767, 6925, 1101, 3612, 12134, 5871, 3640, 860, 4062, 895, 149, 174, 17, 1624, 1846, [187437, 4955, 30920, 6033, 3838, 4332, 3160, 7051, 7677, 9167, 1316, 1372, 833, 6757, 7355, 3302, 3776, 4671,
125, 122, 124, 662, 1452, 697, 222, 190, 786, 200, 450, 295, 79, 205, 156, 181, 70, 64, 337, 1352, 336, 78, 6769, 5706, 3908, 903, 3686, 3596, 6200, 7920, 8779, 4505, 4272, 1862, 4698, 1962, 4403, 6659, 2402, 2689,
628, 841, 287, 83, 702, 1177, 313865, 195, 1081, 882, 1059, 4175, 123, 1700, 2317, 1579, 368, 85]) 4012, 4175, 3411, 17048, 5637, 14553, 3923, 5539, 4289, 10084, 7018, 4314, 3099, 4638, 4939, 5543, 2038, 4004,
5053, 4578, 27292, 4113, 5931, 2905, 11174, 2873, 4036, 3415, 1517, 4122, 1980, 4464, 1190, 2302, 156, 3933,
1877, 17630, 4337, 4624, 1075, 3468, 135, 1380])
weights /= weights.sum() weights /= weights.sum()
return weights[indices] return weights
def plot_one_box(x, img, color=None, label=None, line_thickness=None): # Plots one bounding box on image img def plot_one_box(x, img, color=None, label=None, line_thickness=None): # Plots one bounding box on image img
@ -174,7 +176,7 @@ def bbox_iou(box1, box2, x1y1x2y2=True):
b2_x1, b2_x2 = box2[:, 0] - box2[:, 2] / 2, box2[:, 0] + box2[:, 2] / 2 b2_x1, b2_x2 = box2[:, 0] - box2[:, 2] / 2, box2[:, 0] + box2[:, 2] / 2
b2_y1, b2_y2 = box2[:, 1] - box2[:, 3] / 2, box2[:, 1] + box2[:, 3] / 2 b2_y1, b2_y2 = box2[:, 1] - box2[:, 3] / 2, box2[:, 1] + box2[:, 3] / 2
# get the corrdinates of the intersection rectangle # get the coordinates of the intersection rectangle
inter_rect_x1 = torch.max(b1_x1, b2_x1) inter_rect_x1 = torch.max(b1_x1, b2_x1)
inter_rect_y1 = torch.max(b1_y1, b2_y1) inter_rect_y1 = torch.max(b1_y1, b2_y1)
inter_rect_x2 = torch.min(b1_x2, b2_x2) inter_rect_x2 = torch.min(b1_x2, b2_x2)
@ -192,7 +194,7 @@ def build_targets(pred_boxes, pred_conf, pred_cls, target, anchor_wh, nA, nC, nG
""" """
returns nT, nCorrect, tx, ty, tw, th, tconf, tcls returns nT, nCorrect, tx, ty, tw, th, tconf, tcls
""" """
nB = len(target) # target.shape[0] nB = len(target) # number of images in batch
nT = [len(x) for x in target] # torch.argmin(target[:, :, 4], 1) # targets per image nT = [len(x) for x in target] # torch.argmin(target[:, :, 4], 1) # targets per image
tx = torch.zeros(nB, nA, nG, nG) # batch size (4), number of anchors (3), number of grid points (13) tx = torch.zeros(nB, nA, nG, nG) # batch size (4), number of anchors (3), number of grid points (13)
ty = torch.zeros(nB, nA, nG, nG) ty = torch.zeros(nB, nA, nG, nG)
@ -236,13 +238,6 @@ def build_targets(pred_boxes, pred_conf, pred_cls, target, anchor_wh, nA, nC, nG
u = torch.cat((gi, gj, a), 0).view(3, -1).numpy() u = torch.cat((gi, gj, a), 0).view(3, -1).numpy()
_, first_unique = np.unique(u[:, iou_order], axis=1, return_index=True) # first unique indices _, first_unique = np.unique(u[:, iou_order], axis=1, return_index=True) # first unique indices
# Unique anchor selection (faster but does not retain order) TODO: update to retain original order
# u = gi.float() * 0.4361538773074043 + gj.float() * 0.28012496588736746 + a.float() * 0.6627147212460307
# _, first_unique_sorted = np.unique(u[iou_order], return_index=True) # first unique indices
# Slow - fast difference comparison
# print(((first_unique - first_unique_sorted) ** 2).sum())
i = iou_order[first_unique] i = iou_order[first_unique]
# best anchor must share significant commonality (iou) with target # best anchor must share significant commonality (iou) with target
i = i[iou_anch_best[i] > 0.10] i = i[iou_anch_best[i] > 0.10]
@ -423,7 +418,19 @@ def strip_optimizer_from_checkpoint(filename='checkpoints/best.pt'):
torch.save(a, filename.replace('.pt', '_lite.pt')) torch.save(a, filename.replace('.pt', '_lite.pt'))
def plotResults(): def coco_class_count(path='/Users/glennjocher/downloads/DATA/coco/labels/train2014/'):
import glob
nC = 80 # number classes
x = np.zeros(nC, dtype='int32')
files = sorted(glob.glob('%s/*.*' % path))
for i, file in enumerate(files):
labels = np.loadtxt(file, dtype=np.float32).reshape(-1, 5)
x += np.bincount(labels[:, 0].astype('int32'), minlength=nC)
print(i, len(files))
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