cmc, mAP = eval_sysu(-distmat, query_label, gall_label, query_cam,
gall_cam)
print('Evaluation Time:\t {:.3f}'.format(time.time() - start))
return cmc, mAP
# training
print('==> Start Training...')
for epoch in range(start_epoch, 128 - start_epoch):
print('==> Preparing Data Loader...')
# identity sampler
sampler = IdentitySampler(trainset.train_color_label, \
trainset.train_thermal_label, color_pos, thermal_pos, args.batch_size)
trainset.cIndex = sampler.index1 # color index
trainset.tIndex = sampler.index2 # thermal index
trainloader = data.DataLoader(trainset, batch_size=args.batch_size,\
sampler = sampler, num_workers=args.workers, drop_last =True)
# training
train(epoch)
if epoch > 0 and epoch % 2 == 0:
print('Test Epoch: {}'.format(epoch))
print('Test Epoch: {}'.format(epoch), file=test_log_file)
# testing
cmc, mAP = test(epoch)
print(
'FC: Rank-1: {:.2%} | Rank-5: {:.2%} | Rank-10: {:.2%}| mAP: {:.2%}'
.format(cmc[0], cmc[4], cmc[9], mAP))
writer.add_scalar('mINP_att', mINP_att, epoch)
return cmc, mAP, mINP, cmc_att, mAP_att, mINP_att
# training
print('==> Start Training...')
for epoch in range(start_epoch, 81 - start_epoch):
print('==> Preparing Data Loader...')
# identity sampler:
sampler = IdentitySampler(trainset.train_color_label, \
trainset.train_thermal_label, color_pos, thermal_pos, args.num_pos, args.batch_size,
epoch)
trainset.cIndex = sampler.index1 # color index
trainset.tIndex = sampler.index2 # infrared index
print(epoch)
print(trainset.cIndex)
print(trainset.tIndex)
loader_batch = args.batch_size * args.num_pos
trainloader = data.DataLoader(trainset, batch_size=loader_batch, \
sampler=sampler, num_workers=args.workers, drop_last=True)
# training
wG = train(epoch, wG)
if epoch > 0 and epoch % 2 == 0:
print('Test Epoch: {}'.format(epoch))
print('Test Epoch: {}'.format(epoch), file=test_log_file)