def test_fol(fol_model, test_gen):
'''
Validate future vehicle localization module
Params:
fol_model: The fol model as nn.Module
test_gen: test data generator
Returns:
'''
fol_model.eval() # Sets the module in training mode.
fol_loss = 0
loader = tqdm(test_gen, total=len(test_gen))
FDE = 0
ADE = 0
FIOU = 0
with torch.set_grad_enabled(False):
for batch_idx, data in enumerate(loader):
input_bbox, input_flow, input_ego_motion, target_bbox, target_ego_motion = data
# run forward
fol_predictions,_ = fol_model(input_bbox, input_flow)
# convert to numpy array, use [0] since batchsize if 1 for test
fol_predictions = fol_predictions.to('cpu').numpy()[0]
input_bbox = input_bbox.to('cpu').numpy()[0]
target_bbox = target_bbox.to('cpu').numpy()[0]
# compute FDE, ADE and FIOU metrics used in FVL2019ICRA paper
input_bbox = np.expand_dims(input_bbox, axis=1)
input_bbox = bbox_denormalize(input_bbox, W=1280, H=640)
fol_predictions = bbox_denormalize(fol_predictions, W=1280, H=640)
target_bbox = bbox_denormalize(target_bbox, W=1280, H=640)
fol_predictions_xyxy = cxcywh_to_x1y1x2y2(fol_predictions)
target_bbox_xyxy = cxcywh_to_x1y1x2y2(target_bbox)
ADE += np.mean(np.sqrt(np.sum((target_bbox_xyxy[:,:,:2] - fol_predictions_xyxy[:,:,:2]) ** 2, axis=-1)))
FDE += np.mean(np.sqrt(np.sum((target_bbox_xyxy[:,-1,:2] - fol_predictions_xyxy[:,-1,:2]) ** 2, axis=-1)))
tmp_FIOU = []
for i in range(target_bbox_xyxy.shape[0]):
tmp_FIOU.append(compute_IOU(target_bbox_xyxy[i,-1,:], fol_predictions_xyxy[i,-1,:], format='x1y1x2y2'))
FIOU += np.mean(tmp_FIOU)
print("FDE: %4f; ADE: %4f; FIOU: %4f" % (FDE, ADE, FIOU))
ADE /= len(test_gen.dataset)
FDE /= len(test_gen.dataset)
FIOU /= len(test_gen.dataset)
print("FDE: %4f; ADE: %4f; FIOU: %4f" % (FDE, ADE, FIOU))
def mask_iou_metrics(all_pred_boxes_t,
all_observed_boxes,
W,
H,
multi_box='latest'):
'''
Compute a 0/1 mask for all tracked object and another mask for all observed boxes,
return the difference of the mask as anomaly measures
Params:
all_pred_boxes_t: a dictionary saves all pred_box_t
all_observed_boxes: a dictionary saves all bbox
multi_box: 'latest', 'mverage', 'union'
Return:
mask_iou:
'''
pred_mask = np.zeros([H, W])
observed_mask = np.zeros([H, W])
if len(all_observed_boxes) > 0:
for tarcker_id, bbox in all_observed_boxes.items():
converted_box = cxcywh_to_x1y1x2y2(bbox)
converted_box = bbox_denormalize(converted_box, W=W,
H=H)[0].astype(int)
observed_mask[converted_box[1]:converted_box[3],
converted_box[0]:converted_box[2]] = 1
if len(all_pred_boxes_t) > 0:
for tracker_id, pred_boxes_t in all_pred_boxes_t.items():
if len(pred_boxes_t) <= 0:
continue
if multi_box == 'union':
converted_boxes = cxcywh_to_x1y1x2y2(pred_boxes_t)
converted_boxes = bbox_denormalize(converted_boxes)
converted_boxes = converted_boxes.astype(int)
for box in converted_boxes:
pred_mask[box[1]:box[3], box[0]:box[2]] = 1
elif multi_box == 'latest':
pred_box_t = pred_boxes_t[0:1, :]
converted_box = cxcywh_to_x1y1x2y2(pred_box_t)
converted_box = bbox_denormalize(converted_box, W=W, H=H)
converted_box = converted_box.astype(int)
box = converted_box[0, :]
pred_mask[box[1]:box[3], box[0]:box[2]] = 1
# mask_diff = np.srqrt(np.mean((pred_mask - observed_mask)**2))
mask_iou = compute_mask_iou(np.expand_dims(pred_mask, 0),
np.expand_dims(observed_mask, 0))
return 1 - mask_iou, pred_mask, observed_mask #mask_diff
def boxes_union(boxes):
new_boxes = cxcywh_to_x1y1x2y2(boxes)
xmin = np.min(new_boxes[:, 0])
ymin = np.min(new_boxes[:, 1])
xmax = np.max(new_boxes[:, 2])
ymax = np.max(new_boxes[:, 3])
union = np.sum(boxes[:, 2] * boxes[:, 3])
union_box = np.array([(xmin + xmax) / 2, (ymin + ymax) / 2, xmax - xmin,
ymax - ymin])
return union, union_box
def boxes_union(boxes):
'''
Givent an array where each row is a box (Cx, Cy, w, h)
Return:
intersection: Box intersection area
inter_box: the intersected rectangle box in (cx, cy, w, h)
'''
new_boxes = cxcywh_to_x1y1x2y2(boxes)
xmin = np.min(new_boxes[:, 0])
ymin = np.min(new_boxes[:, 1])
xmax = np.max(new_boxes[:, 2])
ymax = np.max(new_boxes[:, 3])
union = np.sum(boxes[:, 2] * boxes[:, 3])
union_box = np.array([(xmin + xmax) / 2, (ymin + ymax) / 2, xmax - xmin,
ymax - ymin])
return union, union_box
def boxes_intersection(boxes):
boxes = cxcywh_to_x1y1x2y2(boxes)
xmin = np.max(boxes[:, 0])
ymin = np.min(boxes[:, 1])
xmax = np.max(boxes[:, 2])
ymax = np.min(boxes[:, 3])
w_inter = np.max([0, xmax - xmin])
h_inter = np.max([0, ymax - ymin])
intersection = w_inter * h_inter
if intersection == 0:
inter_box = np.array([[0, 0, 0, 0]])
else:
inter_box = np.array([(xmin + xmax) / 2, (ymin + ymax) / 2,
xmax - xmin, ymax - ymin])
return intersection
def boxes_intersection(boxes):
'''
Givent an array where each row is a box (Cx, Cy, w, h)
Return:
intersection: Box intersection area
inter_box: the intersected rectangle box in (cx, cy, w, h)
'''
boxes = cxcywh_to_x1y1x2y2(boxes)
xmin = np.max(boxes[:, 0])
ymin = np.min(boxes[:, 1])
xmax = np.max(boxes[:, 2])
ymax = np.min(boxes[:, 3])
w_inter = np.max([0, xmax - xmin])
h_inter = np.max([0, ymax - ymin])
intersection = w_inter * h_inter
if intersection == 0:
inter_box = np.array([[0, 0, 0, 0]])
else:
inter_box = np.array([(xmin + xmax) / 2, (ymin + ymax) / 2,
xmax - xmin, ymax - ymin])
return intersection