def voc_evaluate(model, data_loader, device):
n_threads = torch.get_num_threads()
torch.set_num_threads(1)
cpu_device = torch.device("cpu")
model.eval()
metric_logger = utils.MetricLogger(delimiter=" ")
header = 'Test:'
all_boxes = [[] for i in range(21)]
image_index = []
for image, targets in metric_logger.log_every(data_loader, 100, header):
image = list(img.to(device) for img in image)
targets = [{k: v.to(device) for k, v in t.items()} for t in targets]
torch.cuda.synchronize()
model_time = time.time()
outputs = model(image)
name = ''.join([chr(i) for i in targets[0]['name'].tolist()])
image_index.append(name)
outputs = [{k: v.to(cpu_device) for k, v in t.items()} for t in outputs]
image_boxes = [[] for i in range(21)]
for o in outputs:
for i in range(o['boxes'].shape[0]):
image_boxes[o['labels'][i]].extend([
torch.cat([o['boxes'][i],o['scores'][i].unsqueeze(0)], dim=0)
])
#makes sure that the all_boxes is filled with empty array when
#there are no boxes in image_boxes
for i in range(21):
if image_boxes[i] != []:
all_boxes[i].append([torch.stack(image_boxes[i])])
else:
all_boxes[i].append([])
model_time = time.time() - model_time
metric_logger.synchronize_between_processes()
all_boxes_gathered = utils.all_gather(all_boxes)
image_index_gathered = utils.all_gather(image_index)
# results from all processes are gathered here
if utils.is_main_process():
all_boxes = [[] for i in range(21)]
for abgs in all_boxes_gathered:
for ab,abg in zip(all_boxes,abgs):
ab += abg
image_index = []
for iig in image_index_gathered:
image_index+=iig
_write_voc_results_file(all_boxes,image_index, data_loader.dataset.root,
data_loader.dataset._transforms.transforms[0].CLASSES)
_do_python_eval(data_loader)
torch.set_num_threads(n_threads)
def _val_step(features, labels, metrics):
val_logits = model(features, training=False)
val_loss = _replica_loss(labels, val_logits)
val_loss = utils.all_reduce(val_loss,
combiner="sum",
comm_options=comm_options)
labels = tf.identity(labels)
val_logits = utils.all_gather(val_logits,
axis=0,
comm_options=comm_options)
labels = utils.all_gather(labels, axis=0, comm_options=comm_options)
return val_logits, labels, val_loss
def merge(img_ids, eval_imgs):
all_img_ids = utils.all_gather(img_ids)
all_eval_imgs = utils.all_gather(eval_imgs)
merged_img_ids = []
for p in all_img_ids:
merged_img_ids.extend(p)
merged_eval_imgs = []
for p in all_eval_imgs:
merged_eval_imgs.append(p)
merged_img_ids = np.array(merged_img_ids)
merged_eval_imgs = np.concatenate(merged_eval_imgs, 2)
# keep only unique (and in sorted order) images
merged_img_ids, idx = np.unique(merged_img_ids, return_index=True)
merged_eval_imgs = merged_eval_imgs[..., idx]
return merged_img_ids, merged_eval_imgs
def accumulate_predictions_from_multiple_gpus(predictions_per_gpu):
all_predictions = all_gather(predictions_per_gpu)
if not is_main_process():
return
# merge the list of dicts
predictions = {}
for p in all_predictions:
predictions.update(p)
# convert a dict where the key is the index in a list
image_ids = list(sorted(predictions.keys()))
if len(image_ids) != image_ids[-1] + 1:
logger = logging.getLogger("RetinaNet.inference")
logger.warning(
"Number of images that were gathered from multiple processes is not "
"a contiguous set. Some images might be missing from the evaluation"
)
# convert to a list
predictions = [predictions[i] for i in image_ids]
return predictions
