def grid_result_eval(predict_path, log_path='grid_eval.log'):
pids4probes = np.genfromtxt(predict_path, delimiter=' ')
probe_shoot = [0, 0, 0, 0, 0]
for i, pids in enumerate(pids4probes):
for j, pid in enumerate(pids):
if pid - i == 775:
if j == 0:
for k in range(5):
probe_shoot[k] += 1
elif j < 5:
for k in range(1, 5):
probe_shoot[k] += 1
elif j < 10:
for k in range(2, 5):
probe_shoot[k] += 1
elif j < 20:
for k in range(3, 5):
probe_shoot[k] += 1
elif j < 50:
for k in range(4, 5):
probe_shoot[k] += 1
break
probe_acc = [shoot / len(pids4probes) for shoot in probe_shoot]
write(log_path, predict_path + '\n')
write(log_path,
'%.2f\t%.2f\t%.2f\n' % (probe_acc[0], probe_acc[1], probe_acc[2]))
print(predict_path)
print(probe_acc)
def market_result_eval(result_argsort, test_info, query_info, log_path='market_result_eval.log'):
#res = np.genfromtxt(predict_path, delimiter=' ')
res = result_argsort
print('start evaluate map and rank acc')
rank1, mAP = map_rank_quick_eval(query_info, test_info, res)
write(log_path, '%f\t%f\n' % (rank1, mAP))
def result_eval(predict_path,
log_path='market_result_eval.log',
TEST='Market-1501/test',
QUERY='Market-1501/probe'):
res = np.genfromtxt(predict_path, delimiter=' ')
print('predict info get, extract gallery info start')
gallery_cam, gallery_label = get_id(TEST)
gallery_cam = np.array(gallery_cam)
gallery_label = np.array(gallery_label)
query_cam, query_label = get_id(QUERY)
CMC = torch.IntTensor(len(gallery_label)).zero_()
ap = 0.0
for i in range(len(query_label)):
ap_tmp, CMC_tmp = evaluate(res[i], query_label[i], query_cam[i],
gallery_label, gallery_cam)
if CMC_tmp[0] == -1:
continue
CMC = CMC + CMC_tmp
ap += ap_tmp
CMC = CMC.float()
CMC = CMC / len(query_label) # average CMC
print(datetime.now().strftime("%Y.%m.%d-%H:%M:%S\t") + predict_path +
'\nRank@1:%f Rank@5:%f Rank@10:%f mAP:%f\n' %
(CMC[0], CMC[4], CMC[9], ap / len(query_label)))
write(
log_path,
datetime.now().strftime("%Y.%m.%d-%H:%M:%S\t") + predict_path +
'\tRank@1 Rank@5 Rank@10 mAP:\n%f\t%f\t%f\t%f\n' %
(CMC[0], CMC[4], CMC[9], ap / len(query_label)))
def market_result_eval(result_argsort,
test_info,
query_info,
log_path='market_result_eval.log'):
res = result_argsort
print('start evaluate map and rank acc')
rank1_acc, rank5_acc, rank10_acc, mAP = map_rank_quick_eval(
query_info, test_info, res)
write(log_path,
'%f\t%f\t%f\t%f\n' % (rank1_acc, rank5_acc, rank10_acc, mAP))
def market_result_eval(predict_path, log_path='market_result_eval.log', TEST='Market-1501/test',
QUERY='Market-1501/probe'):
res = np.genfromtxt(predict_path, delimiter=' ')
print('predict info get, extract gallery info start')
test_info = extract_info(TEST)
print('extract probe info start')
query_info = extract_info(QUERY)
print('start evaluate map and rank acc')
rank1_acc, rank5_acc, rank10_acc, mAP = map_rank_quick_eval(query_info, test_info, res)
write(log_path, predict_path + '\n')
write(log_path, '& %f & %f & %f & %f\n' % (rank1_acc, rank5_acc, rank10_acc, mAP))
def market_result_eval(sort_gallery_idx,
log_path='market_result_eval.log',
TEST='Market-1501/test',
QUERY='Market-1501/probe'):
print('predict info get, extract gallery info start')
test_info = extract_info(TEST)
print('extract probe info start')
query_info = extract_info(QUERY)
print('start evaluate map and rank acc')
rank1, rank5, rank10, mAP = map_rank_quick_eval(query_info, test_info,
sort_gallery_idx)
write(log_path, '%f\t%f\t%f\t%f\n' % (rank1, rank5, rank10, mAP))
query_feature = query_feature.cuda()
gallery_feature = gallery_feature.cuda()
# print(query_feature.shape)
CMC = torch.IntTensor(len(gallery_label)).zero_()
ap = 0.0
#print(query_label)
scores, indexs = [], []
for i in range(len(query_label)):
ap_tmp, CMC_tmp, index, score = evaluate(query_feature[i],query_label[i],query_cam[i],gallery_feature,gallery_label,gallery_cam)
if CMC_tmp[0]==-1:
continue
CMC = CMC + CMC_tmp
ap += ap_tmp
scores.append(score)
indexs.append(index)
# print(i, CMC_tmp[0])
scores = np.array(scores)
indexs = np.array(indexs)
score_path = os.path.join('eval', opt.name, 'score.txt')
pid_path = os.path.join('eval', opt.name, 'pid.txt')
np.savetxt(score_path, scores, fmt='%.4f')
np.savetxt(pid_path, indexs, fmt='%d')
CMC = CMC.float()
CMC = CMC/len(query_label) #average CMC
print(datetime.now().strftime("%Y.%m.%d-%H:%M:%S\t") + 'Rank@1:%f Rank@5:%f Rank@10:%f mAP:%f\n'%(CMC[0],CMC[4],CMC[9],ap/len(query_label)))
write(os.path.join('eval', opt.name, 'acc.txt'),
datetime.now().strftime("%Y.%m.%d-%H:%M:%S\t") + 'Rank@1:%f Rank@5:%f Rank@10:%f mAP:%f\n'%(CMC[0],CMC[4],CMC[9],ap/len(query_label)))
y = model.predict_generator(pig_generator(test_imgs, batch_size),
3000 / batch_size + 1,
use_multiprocessing=True)
# ys = np.sum(y,axis=1)
# y_i = ys > 0
y /= 1.001
predict_path = 'predict.csv'
if path.exists(predict_path):
remove(predict_path)
np.savetxt(predict_path, y, fmt='%6f', delimiter='\t')
# y = np.genfromtxt('predict.csv', delimiter='\t')
y = y.reshape(-1)
ids = np.array(test_ids)
cls = np.arange(1, 31)
ids = ids.repeat(30)
cls = np.tile(cls, 3000)
rst = np.append(ids, cls)
rst = np.append(rst, y).reshape(3, 90000)
rst = np.rot90(rst)
rst_str = ''
for line in rst:
rst_str += '%d,%d,%5f\n' % (int(line[0]), int(line[1]), float(line[2]))
rst_path = 'pig_rst.csv'
if path.exists(rst_path):
remove(rst_path)
write(rst_path, rst_str)
# np.savetxt('pig_rst.csv', rst, fmt='%d,%d,%6f', delimiter=',')