#!/usr/local/bin/python3
# -*- coding: utf-8 -*-
import os
import getpass
import argparse
import logging
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import torch
import torch.nn as nn
from torch.utils.data import DataLoader
from torchvision import transforms
from dataset.lip import LIP
from net.pspnet import PSPNet
models = {
'squeezenet': lambda: PSPNet(sizes=(1, 2, 3, 6), psp_size=512, deep_features_size=256, backend='squeezenet'),
'densenet': lambda: PSPNet(sizes=(1, 2, 3, 6), psp_size=1024, deep_features_size=512, backend='densenet'),
'resnet18': lambda: PSPNet(sizes=(1, 2, 3, 6), psp_size=512, deep_features_size=256, backend='resnet18'),
'resnet34': lambda: PSPNet(sizes=(1, 2, 3, 6), psp_size=512, deep_features_size=256, backend='resnet34'),
'resnet50': lambda: PSPNet(sizes=(1, 2, 3, 6), psp_size=2048, deep_features_size=1024, backend='resnet50'),
'resnet101': lambda: PSPNet(sizes=(1, 2, 3, 6), psp_size=2048, deep_features_size=1024, backend='resnet101'),
'resnet152': lambda: PSPNet(sizes=(1, 2, 3, 6), psp_size=2048, deep_features_size=1024, backend='resnet152')
}
parser = argparse.ArgumentParser(description="Pyramid Scene Parsing Network")
parser.add_argument('--data-path', type=str, help='Path to dataset folder')
parser.add_argument('--models-path', type=str, default='./checkpoints', help='Path for storing model snapshots')
parser.add_argument('--backend', type=str, default='densenet', help='Feature extractor')
parser.add_argument('--gpu', type=str, default='0', help='List of GPUs for parallel training, e.g. 0,1,2,3')
parser.add_argument('--batch-size', type=int, default=1, help="Number of images sent to the network in one step.")
parser.add_argument('--num-classes', type=int, default=20, help="Number of classes.")
parser.add_argument('-v', '--visualize', action='store_true', help="Display output and ground truth.")
args = parser.parse_args()
def build_network(snapshot, backend):
epoch = 0
backend = backend.lower()
net = models[backend]()
net = nn.DataParallel(net)
if snapshot is not None:
_, epoch = os.path.basename(snapshot).split('_')
if not epoch == 'last':
epoch = int(epoch)
net.load_state_dict(torch.load(snapshot))
logging.info("Snapshot for epoch {} loaded from {}".format(epoch, snapshot))
net = net.cuda()
return net, epoch
def get_transform():
transform_image_list = [
transforms.Resize((256, 256), 3),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
]
transform_gt_list = [
transforms.Resize((256, 256), 0),
transforms.Lambda(lambda img: np.asarray(img, dtype=np.uint8)),
]
data_transforms = {
'img': transforms.Compose(transform_image_list),
'gt': transforms.Compose(transform_gt_list),
}
return data_transforms
def show_image(img, pred, gt):
fig, axes = plt.subplots(1, 3)
ax0, ax1, ax2 = axes
ax0.get_xaxis().set_ticks([])
ax0.get_yaxis().set_ticks([])
ax1.get_xaxis().set_ticks([])
ax1.get_yaxis().set_ticks([])
ax2.get_xaxis().set_ticks([])
ax2.get_yaxis().set_ticks([])
classes = np.array(('Background', # always index 0
'Hat', 'Hair', 'Glove', 'Sunglasses',
'UpperClothes', 'Dress', 'Coat', 'Socks',
'Pants', 'Jumpsuits', 'Scarf', 'Skirt',
'Face', 'Left-arm', 'Right-arm', 'Left-leg',
'Right-leg', 'Left-shoe', 'Right-shoe', ))
colormap = [(0,0,0),
(1,0.25,0), (0,0.25,0), (0.5,0,0.25), (1,1,1),
(1,0.75,0), (0,0,0.5), (0.5,0.25,0), (0.75,0,0.25),
(1,0,0.25), (0,0.5,0), (0.5,0.5,0), (0.25,0,0.5),
(1,0,0.75), (0,0.5,0.5), (0.25,0.5,0.5), (1,0,0),
(1,0.25,0), (0,0.75,0), (0.5,0.75,0), ]
cmap = matplotlib.colors.ListedColormap(colormap)
bounds=[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20]
norm = matplotlib.colors.BoundaryNorm(bounds, cmap.N)
h, w, _ = pred.shape
def denormalize(img, mean, std):
c, _, _ = img.shape
for idx in range(c):
img[idx, :, :] = img[idx, :, :] * std[idx] + mean[idx]
return img
img = denormalize(img[0].numpy(), [0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
img = img.transpose(1,2,0).reshape((h,w,3))
pred = pred.reshape((h,w))
gt = gt.reshape((h, w))
# show image
ax0.set_title('img')
ax0.imshow(img)
ax1.set_title('pred')
mappable = ax1.imshow(pred, cmap=cmap, norm=norm)
ax2.set_title('gt')
mappable = ax2.imshow(gt, cmap=cmap, norm=norm)
# colorbar legend
cbar = plt.colorbar(mappable, ax=axes, shrink=0.7,) # orientation='horizontal'
cbar.ax.get_yaxis().set_ticks([])
for j, lab in enumerate(classes):
cbar.ax.text(1.3, (j+0.45) / 20.0, lab, ha='left', va='center',) # fontsize=7
# cbar.ax.get_yaxis().labelpad = 5
plt.savefig(fname="result.jpg")
plt.show()
def get_pixel_acc(pred, gt):
valid = (gt >= 0)
acc_sum = (valid * (pred == gt)).sum()
valid_sum = valid.sum()
acc = float(acc_sum) / (valid_sum + 1e-10)
return acc
def get_mean_acc(pred, gt, numClass):
imPred = pred.copy()
imLabel = gt.copy()
imPred += 1
imLabel += 1
imPred = imPred * (imLabel > 0)
# Compute area intersection:
intersection = imPred * (imPred == imLabel)
(area_intersection, _) = np.histogram(
intersection, bins=numClass, range=(1, numClass))
# Compute area label:
(area_label, _) = np.histogram(imLabel, bins=numClass, range=(1, numClass))
# Remove classes from unlabeled pixels in gt image.
# We should not penalize detections in unlabeled portions of the image.
valid = area_label > 0
# Compute intersection over union:
classes_acc = area_intersection / (area_label + 1e-10)
mean_acc = np.average(classes_acc, weights=valid)
return mean_acc
def get_mean_IoU(pred, gt, numClass):
imPred = pred.copy()
imLabel = gt.copy()
imPred += 1
imLabel += 1
imPred = imPred * (imLabel > 0)
# Compute area intersection:
intersection = imPred * (imPred == imLabel)
(area_intersection, _) = np.histogram(
intersection, bins=numClass, range=(1, numClass))
# Compute area union:
(area_pred, _) = np.histogram(imPred, bins=numClass, range=(1, numClass))
(area_label, _) = np.histogram(imLabel, bins=numClass, range=(1, numClass))
area_union = area_pred + area_label - area_intersection
# Remove classes from unlabeled pixels in gt image.
# We should not penalize detections in unlabeled portions of the image.
valid = area_label > 0
# Compute intersection over union:
IoU = area_intersection / (area_union + 1e-10)
mean_IoU = np.average(IoU, weights=valid)
return mean_IoU
def get_mean_acc_and_IoU(pred, gt, numClass):
imPred = pred.copy()
imLabel = gt.copy()
imPred += 1
imLabel += 1
imPred = imPred * (imLabel > 0)
# Compute area intersection:
intersection = imPred * (imPred == imLabel)
(area_intersection, _) = np.histogram(
intersection, bins=numClass, range=(1, numClass))
# Compute area union:
(area_pred, _) = np.histogram(imPred, bins=numClass, range=(1, numClass))
(area_label, _) = np.histogram(imLabel, bins=numClass, range=(1, numClass))
area_union = area_pred + area_label - area_intersection
# Remove classes from unlabeled pixels in gt image.
# We should not penalize detections in unlabeled portions of the image.
valid = area_label > 0
# Compute mean acc.
classes_acc = area_intersection / (area_label + 1e-10)
mean_acc = np.average(classes_acc, weights=valid)
# Compute intersection over union:
IoU = area_intersection / (area_union + 1e-10)
mean_IoU = np.average(IoU, weights=valid)
return mean_acc, mean_IoU
def main():
# --------------- model --------------- #
snapshot = os.path.join(args.models_path, args.backend, 'PSPNet_last')
net, starting_epoch = build_network(snapshot, args.backend)
net.eval()
# ------------ data loader ------------ #
data_transform = get_transform()
val_loader = DataLoader(LIP(args.data_path, train=False, transform=data_transform['img'],
gt_transform=data_transform['gt']),
batch_size=args.batch_size,
shuffle=False,
)
# --------------- eval --------------- #
overall_acc_list = []
mean_acc_list = []
mean_IoU_list = []
with torch.no_grad():
for index, (img, gt) in enumerate(val_loader):
pred_seg, pred_cls = net(img.cuda())
pred_seg = pred_seg[0]
pred = pred_seg.cpu().numpy().transpose(1, 2, 0)
pred = np.asarray(np.argmax(pred, axis=2), dtype=np.uint8).reshape((256, 256, 1))
gt = np.asarray(gt.numpy(), dtype=np.uint8).transpose(1, 2, 0)
if args.visualize:
show_image(img, pred, gt)
overall_acc_list.append(get_pixel_acc(pred, gt))
_mean_acc, _mean_IoU = get_mean_acc_and_IoU(pred, gt, args.num_classes)
mean_acc_list.append(_mean_acc)
mean_IoU_list.append(_mean_IoU)
print(' %d / %d ' % (index, len(val_loader)))
print(' overall acc. : %f ' % (np.mean(overall_acc_list)))
print(' mean acc. : %f ' % (np.mean(mean_acc_list)))
print(' mean IoU : %f ' % (np.mean(mean_IoU_list)))
if __name__ == '__main__':
main()
