keypoint tensorboard

models/keypoint/trainer.py

@@ -180,30 +180,30 @@ def evaluate(model, data_loader, epoch, writer, device):
 
         model_time = time.time()
         outputs = model(images)
-        print(f'outputs:{outputs}')
+        # print(f'outputs:{outputs}')
 
         if batch_idx == 0:
             show_line(images[0], outputs[0], epoch, writer)
 
-        outputs = [{k: v.to(cpu_device) for k, v in t.items()} for t in outputs]
-        model_time = time.time() - model_time
-
-        res = {target["image_id"]: output for target, output in zip(targets, outputs)}
-        evaluator_time = time.time()
-        coco_evaluator.update(res)
-        evaluator_time = time.time() - evaluator_time
-        metric_logger.update(model_time=model_time, evaluator_time=evaluator_time)
-
-    # gather the stats from all processes
-    metric_logger.synchronize_between_processes()
-    print("Averaged stats:", metric_logger)
-    coco_evaluator.synchronize_between_processes()
-
-    # accumulate predictions from all images
-    coco_evaluator.accumulate()
-    coco_evaluator.summarize()
-    torch.set_num_threads(n_threads)
-    return coco_evaluator
+    #     outputs = [{k: v.to(cpu_device) for k, v in t.items()} for t in outputs]
+    #     #     model_time = time.time() - model_time
+    #     #
+    #     #     res = {target["image_id"]: output for target, output in zip(targets, outputs)}
+    #     #     evaluator_time = time.time()
+    #     #     coco_evaluator.update(res)
+    #     #     evaluator_time = time.time() - evaluator_time
+    #     #     metric_logger.update(model_time=model_time, evaluator_time=evaluator_time)
+    #     #
+    #     # # gather the stats from all processes
+    #     # metric_logger.synchronize_between_processes()
+    #     # print("Averaged stats:", metric_logger)
+    #     # coco_evaluator.synchronize_between_processes()
+    #     #
+    #     # # accumulate predictions from all images
+    #     # coco_evaluator.accumulate()
+    #     # coco_evaluator.summarize()
+    #     # torch.set_num_threads(n_threads)
+    #     # return coco_evaluator
 
 
 def train_cfg(model, cfg):
@@ -352,7 +352,15 @@ def get_transform(is_train, **kwargs):
 def write_metric_logs(epoch, metric_logger, writer):
     writer.add_scalar(f'loss_classifier:', metric_logger.meters['loss_classifier'].global_avg, epoch)
     writer.add_scalar(f'loss_box_reg:', metric_logger.meters['loss_box_reg'].global_avg, epoch)
-    writer.add_scalar(f'loss_mask:', metric_logger.meters['loss_mask'].global_avg, epoch)
+    # writer.add_scalar(f'loss_mask:', metric_logger.meters['loss_mask'].global_avg, epoch)
+    writer.add_scalar('Loss/box_reg', metric_logger.meters['loss_keypoint'].global_avg, epoch)
     writer.add_scalar(f'loss_objectness:', metric_logger.meters['loss_objectness'].global_avg, epoch)
     writer.add_scalar(f'loss_rpn_box_reg:', metric_logger.meters['loss_rpn_box_reg'].global_avg, epoch)
     writer.add_scalar(f'train loss:', metric_logger.meters['loss'].global_avg, epoch)
+
+# def log_losses_to_tensorboard(writer, result, step):
+#     writer.add_scalar('Loss/classifier', result['loss_classifier'].item(), step)
+#     writer.add_scalar('Loss/box_reg', result['loss_box_reg'].item(), step)
+#     writer.add_scalar('Loss/box_reg', result['loss_keypoint'].item(), step)
+#     writer.add_scalar('Loss/objectness', result['loss_objectness'].item(), step)
+#     writer.add_scalar('Loss/rpn_box_reg', result['loss_rpn_box_reg'].item(), step)