Merge remote-tracking branch 'OpenPCDet-fnozarian/DA-52-cos-score-wei…

…ghting-pooled-refactor' into DA-52-cos-score-weighting-pooled-refactor
fnozarian · Sep 3, 2023 · 0685d17 · 0685d17
2 parents 87bfa2c + 3889d9d
commit 0685d17
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 3 deletions.
diff --git a/pcdet/models/roi_heads/roi_head_template.py b/pcdet/models/roi_heads/roi_head_template.py
@@ -119,7 +119,7 @@ def proposal_layer(self, batch_dict,  nms_config):
         rois = batch_box_preds.new_zeros((batch_size, nms_config.NMS_POST_MAXSIZE, batch_box_preds.shape[-1]))
         roi_scores = batch_box_preds.new_zeros((batch_size, nms_config.NMS_POST_MAXSIZE))
         roi_labels = batch_box_preds.new_zeros((batch_size, nms_config.NMS_POST_MAXSIZE), dtype=torch.long)
-
+        roi_scores_multiclass = batch_box_preds.new_zeros((batch_size, nms_config.NMS_POST_MAXSIZE, batch_cls_preds.shape[-1]))
         for index in range(batch_size):
             if batch_dict.get('batch_index', None) is not None:
                 assert batch_cls_preds.shape.__len__() == 2
@@ -142,9 +142,11 @@ def proposal_layer(self, batch_dict,  nms_config):
             rois[index, :len(selected), :] = box_preds[selected]
             roi_scores[index, :len(selected)] = cur_roi_scores[selected]
             roi_labels[index, :len(selected)] = cur_roi_labels[selected]
+            roi_scores_multiclass[index, :len(selected), :] = cls_preds[selected]
 
         batch_dict['rois'] = rois
         batch_dict['roi_scores'] = roi_scores
+        batch_dict['roi_scores_multiclass'] = roi_scores_multiclass
         batch_dict['roi_labels'] = roi_labels + 1
         batch_dict['has_class_labels'] = True if batch_cls_preds.shape[-1] > 1 else False
         batch_dict.pop('batch_index', None)

diff --git a/tools/cfgs/kitti_models/pv_rcnn_ssl_60.yaml b/tools/cfgs/kitti_models/pv_rcnn_ssl_60.yaml
@@ -162,7 +162,7 @@ MODEL:
         ENABLE_VIS: False
         ENABLE_PROTO_CONTRASTIVE_LOSS: False
         PROTO_CONTRASTIVE_LOSS_WEIGHT: 1.0
-        ENABLE_SOFT_TEACHER: True
+        ENABLE_SOFT_TEACHER: False
         ENABLE_ULB_CLS_DIST_LOSS: False
         ENABLE_EVAL: True
         METRICS_PRED_TYPES: [roi_pl_gt, pl_gt_metrics_before_filtering]
@@ -221,7 +221,7 @@ MODEL:
 
             REG_FG_THRESH: 0.55
             UNLABELED_REG_FG_THRESH: [0.55, 0.55, 0.55]
-            UNLABELED_SAMPLER_TYPE: subsample_unlabeled_rois_default
+            UNLABELED_SAMPLER_TYPE: subsample_labeled_rois #subsample_unlabeled_rois_default
             UNLABELED_SAMPLE_EASY_BG: False
             UNLABELED_SHARP_RCNN_CLS_LABELS: True
             UNLABELED_USE_CALIBRATED_IOUS: True