def offline_evaluate(pred_func, output_file):
df = get_test_dataflow(add_mask=False)
all_results = eval_on_dataflow(
df, lambda img: detect_one_image(img, pred_func))
#print(all_results)
# input()
with open(output_file, 'w') as f:
json.dump(all_results, f, cls=MyEncoder)
ret = print_evaluation_scores(output_file)
print(ret)
def visualize(model_path, nr_visualize=50, output_dir='visualize'):
df = get_test_dataflow() # we don't visualize mask stuff
df.reset_state()
pred = OfflinePredictor(
PredictConfig(model=Model(),
session_init=get_model_loader(model_path),
input_names=['image', 'gt_boxes', 'gt_labels'],
output_names=[
'generate_rpn_proposals/boxes',
'generate_rpn_proposals/probs',
'fastrcnn_all_probs',
'final_boxes',
'final_probs',
'final_labels',
'final_masks',
]))
if os.path.isdir(output_dir):
shutil.rmtree(output_dir)
utils.fs.mkdir_p(output_dir)
with tqdm.tqdm(total=nr_visualize) as pbar:
for idx, dp in itertools.islice(enumerate(df.get_data()),
nr_visualize):
img, _, _, gt_boxes, gt_labels, _ = dp
rpn_boxes, rpn_scores, all_probs, \
final_boxes, final_probs, final_labels, final_masks = pred(img, gt_boxes, gt_labels)
# draw groundtruth boxes
gt_viz = draw_annotation(img, gt_boxes, gt_labels)
# draw best proposals for each groundtruth, to show recall
proposal_viz, good_proposals_ind = draw_proposal_recall(
img, rpn_boxes, rpn_scores, gt_boxes)
# draw the scores for the above proposals
score_viz = draw_predictions(img, rpn_boxes[good_proposals_ind],
all_probs[good_proposals_ind])
#results = [DetectionResult(*args) for args in zip(final_boxes, final_probs, final_labels, final_masks)]
results = [
DetectionResult(*args)
for args in zip(final_boxes, final_probs, final_labels,
[None] * len(final_labels))
]
final_viz = draw_final_outputs(img, results)
viz = tpviz.stack_patches(
[gt_viz, proposal_viz, score_viz, final_viz], 2, 2)
# if os.environ.get('DISPLAY', None):
# tpviz.interactive_imshow(viz)
cv2.imwrite("{}/{:03d}.png".format(output_dir, idx), viz)
pbar.update()
def offline_evaluate(pred_func, output_file):
df = get_test_dataflow()
all_results = eval_on_dataflow(
df, lambda img: detect_one_image(img, pred_func))
print(all_results)
def _setup_graph(self):
self.pred = self.trainer.get_predictor(['image'],
get_model_output_names())
self.df = get_test_dataflow(add_mask=True)
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
if args.visualize or args.evaluate or args.predict:
# autotune is too slow for inference
os.environ['TF_CUDNN_USE_AUTOTUNE'] = '0'
assert args.load
print_config()
if args.evaluate:
pred = OfflinePredictor(
PredictConfig(model=Model(),
session_init=get_model_loader(args.load),
input_names=['image'],
output_names=get_model_output_names()))
df = get_test_dataflow(add_mask=True)
df.reset_state()
all_results, local_score = eval_on_dataflow(
df, lambda img: detect_one_image(img, pred))
print("F2 Score: ", local_score)
elif args.predict:
imgs = Detection.load_many(
config.BASEDIR, config.TEST_DATASET,
add_gt=False) # to load the class names into caches
# filter with zero-ship
imgs = [(img['image_data'], img['id']) for img in imgs]
pred = OfflinePredictor(
PredictConfig(model=Model(),
session_init=get_model_loader(args.load),
input_names=['image'],