deep-person-reid/eval_metrics.py
2018-03-11 21:22:14 +00:00

52 lines
1.5 KiB
Python

from __future__ import absolute_import
import numpy as np
import copy
def evaluate(distmat, q_pids, g_pids, q_camids, g_camids):
num_q = distmat.shape[0]
indices = np.argsort(distmat, axis=1)
matches = (g_pids[indices] == q_pids[:, np.newaxis]).astype(np.int32)
# compute cmc curve for each query, maximum rank is fixed to _MAX_RANK
all_cmc = []
all_AP = []
num_valid_q = 0.
for q_idx in range(num_q):
# get query pid and camid
q_pid = q_pids[q_idx]
q_camid = q_camids[q_idx]
# remove gallery samples that have the same pid and camid with query
order = indices[q_idx]
remove = (g_pids[order] == q_pid) and (g_camids[order] == q_camid)
keep = np.invert(remove)
# compute cmc curve
cmc = matches[q_idx][keep]
if not np.any(cmc):
# this condition is true when query identity does not appear in gallery
continue
cmc = cmc.cumsum()
cmc[cmc > 1] = 1
all_cmc.append(cmc)
num_valid_q += 1.
num_rel = cmc.sum()
tmp_cmc = cmc.cumsum()
tmp_cmc = [x / (i+1.) for i, x in enumerate(tmp_cmc)]
tmp_cmc = np.asarray(tmp_cmc)
AP = tmp_cmc.sum() / num_rel
all_AP.append(AP)
assert num_valid_q > 0, "Error: All query identities do not appear in gallery"
all_cmc = np.asarray(all_cmc).astype(np.float32)
all_cmc = all_cmc.sum(0) / num_valid_q
mAP = np.mean(all_AP)
return all_cmc, mAP