def eval_revop(p):
# revisited evaluation
gnd = cfg['gnd']
# evaluate ranks
ks = [1, 5, 10]
# search for easy & hard
gnd_t = []
for i in range(len(gnd)):
g = {}
g['ok'] = np.concatenate([gnd[i]['easy'], gnd[i]['hard']])
g['junk'] = np.concatenate([gnd[i]['junk']])
gnd_t.append(g)
mapM, _, _, _ = compute_map(p, gnd_t, ks)
# search for hard
gnd_t = []
for i in range(len(gnd)):
g = {}
g['ok'] = np.concatenate([gnd[i]['hard']])
g['junk'] = np.concatenate([gnd[i]['junk'], gnd[i]['easy']])
gnd_t.append(g)
mapH, _, _, _ = compute_map(p, gnd_t, ks)
return np.around(mapH*100, decimals=2),np.around(mapM*100, decimals=2)
def eval_revop(p, silent=False):
# revisited evaluation
gnd = cfg['gnd']
# evaluate ranks
ks = [1, 5, 10]
# search for easy
gnd_t = []
for i in range(len(gnd)):
g = {}
g['ok'] = np.concatenate([gnd[i]['easy']])
g['junk'] = np.concatenate([gnd[i]['junk'], gnd[i]['hard']])
gnd_t.append(g)
mapE, apsE, mprE, prsE = compute_map(p, gnd_t, ks)
# search for easy & hard
gnd_t = []
for i in range(len(gnd)):
g = {}
g['ok'] = np.concatenate([gnd[i]['easy'], gnd[i]['hard']])
g['junk'] = np.concatenate([gnd[i]['junk']])
gnd_t.append(g)
mapM, apsM, mprM, prsM = compute_map(p, gnd_t, ks)
# search for hard
gnd_t = []
for i in range(len(gnd)):
g = {}
g['ok'] = np.concatenate([gnd[i]['hard']])
g['junk'] = np.concatenate([gnd[i]['junk'], gnd[i]['easy']])
gnd_t.append(g)
mapH, apsH, mprH, prsH = compute_map(p, gnd_t, ks)
if not silent:
print('>> {}: mAP E: {}, M: {}, H: {}'.format(
test_dataset, np.around(mapE * 100, decimals=2),
np.around(mapM * 100, decimals=2), np.around(mapH * 100,
decimals=2)))
print('>> {}: mP@k{} E: {}, M: {}, H: {}'.format(
test_dataset, np.array(ks), np.around(mprE * 100, decimals=2),
np.around(mprM * 100, decimals=2), np.around(mprH * 100,
decimals=2)))
return np.around(mapH * 100, decimals=2)
def simgle_img_map(model, output, labels):
actual_num_labels = 0
all_detections = [[[
np.array([]) for _ in range(model.params['num_classes'])
]]]
# n our model if no results it outputs int 0
if output is not 0:
# Get predicted boxes, confidence scores and labels
pred_boxes = output[:, 1:6].cpu().numpy()
scores = output[:, 5].cpu().numpy()
pred_labels = output[:, 7].cpu().numpy()
# Order by confidence
sort_i = np.argsort(scores)
pred_labels = pred_labels[sort_i]
pred_boxes = pred_boxes[sort_i]
for c in range(model.params['num_classes']):
all_detections[0][-1][c] = pred_boxes[pred_labels == c]
all_annotations = []
for label_ in labels:
all_annotations.append(
[np.array([]) for _ in range(model.params['num_classes'])])
if any(label_[:, -1] > 0):
annotation_labels = label_[label_[:, -1] > 0, 0]
_annotation_boxes = label_[label_[:, -1] > 0, 1:]
num_labels = len(np.unique(annotation_labels))
# Reformat to x1, y1, x2, y2 and rescale to image dim
annotation_boxes = np.empty_like(_annotation_boxes)
annotation_boxes[:, 0] = _annotation_boxes[:, 0] -\
_annotation_boxes[:, 2] / 2
annotation_boxes[:, 1] = _annotation_boxes[:, 1] -\
_annotation_boxes[:, 3] / 2
annotation_boxes[:, 2] = _annotation_boxes[:, 0] +\
_annotation_boxes[:, 2] / 2
annotation_boxes[:, 3] = _annotation_boxes[:, 1] +\
_annotation_boxes[:, 3] / 2
annotation_boxes *= model.params['height']
for label in range(model.params['num_classes']):
all_annotations[-1][label] =\
annotation_boxes[annotation_labels == label, :]
# if train it results consists mAP, average_precisions map_frame
# else: mAP, average_precisions
actual_num_labels = np.max([actual_num_labels, num_labels])
print(f"actual_num_labels : {actual_num_labels}")
mAP, ap = compute_map(all_detections,
all_annotations,
conf_index=0,
map_frame=None,
train=True,
actual_num_labels=actual_num_labels,
params=model.params)
return mAP, ap
ranks = np.argsort(-sim, axis=0)
# revisited evaluation
gnd = cfg['gnd']
# evaluate ranks
ks = [1, 5, 10]
# search for easy
gnd_t = []
for i in range(len(gnd)):
g = {}
g['ok'] = np.concatenate([gnd[i]['easy']])
g['junk'] = np.concatenate([gnd[i]['junk'], gnd[i]['hard']])
gnd_t.append(g)
mapE, apsE, mprE, prsE = compute_map(ranks, gnd_t, ks)
# search for easy & hard
gnd_t = []
for i in range(len(gnd)):
g = {}
g['ok'] = np.concatenate([gnd[i]['easy'], gnd[i]['hard']])
g['junk'] = np.concatenate([gnd[i]['junk']])
gnd_t.append(g)
mapM, apsM, mprM, prsM = compute_map(ranks, gnd_t, ks)
# search for hard
gnd_t = []
for i in range(len(gnd)):
g = {}
g['ok'] = np.concatenate([gnd[i]['hard']])