Fixing config names (MEGAPOSE_DATA_DIR --> HAPPYPOSE_DATA_DIR) and fi…

…xing bugs introduced in the previous merge

README.md

@@ -40,7 +40,7 @@ pip install -r requirements.txt -e .
 
 ```
 Create data dir /somewhere/convenient. The dataset to store are quite large.
-export MEGAPOSE_DATA_DIR=/somewhere/convenient
+export HAPPYPOSE_DATA_DIR=/somewhere/convenient
 ```
 
 # Configuration for the evaluation

happypose/pose_estimators/cosypose/cosypose/config.py

@@ -15,7 +15,7 @@
 PROJECT_ROOT = Path(happypose.__file__).parent.parent
 PROJECT_DIR = PROJECT_ROOT
 DATA_DIR = PROJECT_DIR / 'data'
-LOCAL_DATA_DIR = Path(os.environ.get("MEGAPOSE_DATA_DIR", Path(PROJECT_DIR) / "local_data"))
+LOCAL_DATA_DIR = Path(os.environ.get("HAPPYPOSE_DATA_DIR", Path(PROJECT_DIR) / "local_data"))
 TEST_DATA_DIR = LOCAL_DATA_DIR
 DASK_LOGS_DIR = LOCAL_DATA_DIR / 'dasklogs'
 SYNT_DS_DIR = LOCAL_DATA_DIR / 'synt_datasets'

happypose/pose_estimators/megapose/src/megapose/config.py

@@ -28,7 +28,7 @@
 
 PROJECT_ROOT = Path(megapose.__file__).parent.parent.parent
 PROJECT_DIR = PROJECT_ROOT
-LOCAL_DATA_DIR = Path(os.environ.get("MEGAPOSE_DATA_DIR", Path(PROJECT_DIR) / "local_data"))
+LOCAL_DATA_DIR = Path(os.environ.get("HAPPYPOSE_DATA_DIR", Path(PROJECT_DIR) / "local_data"))
 BOP_DS_DIR = LOCAL_DATA_DIR / "bop_datasets"
 NB_DATA_DIR = LOCAL_DATA_DIR / "notebook_data"
 SHAPENET_DIR = LOCAL_DATA_DIR / "shapenetcorev2"

happypose/pose_estimators/megapose/src/megapose/evaluation/prediction_runner.py

@@ -60,7 +60,6 @@
 
 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
-<<<<<<< HEAD
 
 
 ##################################
@@ -112,8 +111,6 @@
 ##################################
 
 
-=======
->>>>>>> c6cd60e ( first attempt at bop challenge)
 
 class PredictionRunner:
     def __init__(
@@ -161,14 +158,10 @@ def run_inference_pipeline(
 
 
         """
-        print("gt detections =\n", gt_detections)
-        print("sam detections =\n", sam_detections)
-
         # TODO: this check could be done outside of run_inference_pipeline
         # and then only check if detections are None
         if self.inference_cfg.detection_type == "gt":
             detections = gt_detections
-            print("gt detections =", gt_detections.bboxes)
             run_detector = False
         elif self.inference_cfg.detection_type == "sam":
             # print("sam_detections =", sam_detections.bboxes)
@@ -177,14 +170,7 @@ def run_inference_pipeline(
         elif self.inference_cfg.detection_type == "detector":
             detections = None
             run_detector = True
-<<<<<<< HEAD
 
-=======
-        elif self.inference_cfg.detection_type == "sam":
-            print("sam_detections =", sam_detections.bboxes)
-            detections = sam_detections
-            run_detector = False
->>>>>>> c6cd60e ( first attempt at bop challenge)
         else:
             raise ValueError(f"Unknown detection type {self.inference_cfg.detection_type}")
 
@@ -252,7 +238,6 @@ def get_predictions(self, pose_estimator: PoseEstimator) -> Dict[str, PoseEstima
 
         predictions_list = defaultdict(list)
 
-<<<<<<< HEAD
         ######
         # This section opens the detections stored in "baseline.json"
         # format it and store it in a dataframe that will be accessed later
@@ -291,24 +276,12 @@ def get_predictions(self, pose_estimator: PoseEstimator) -> Dict[str, PoseEstima
             df_all_dets = pd.DataFrame.from_records(dets_lst)
 
             df_targets = pd.read_json(self.scene_ds.ds_dir / "test_targets_bop19.json")
-=======
-        # Temporary solution
-        if self.inference_cfg.detection_type == "sam":
-            data_path = Path("/home/emaitre/local_data/bop23/baseline/ycbv/baseline.json")
-            object_data = json.loads(data_path.read_text())
-            for object in object_data:
-                object['bbox'] = [float(i) for i in object['bbox']]
-                object['bbox_modal'] = object['bbox']
-                object['label'] = "ycbv-obj_{}".format(str(object['category_id']).zfill(6))
-            object_data = pd.DataFrame.from_records(object_data)
->>>>>>> c6cd60e ( first attempt at bop challenge)
 
         for n, data in enumerate(tqdm(self.dataloader)):
             # data is a dict
             rgb = data["rgb"]
             depth = data["depth"]
             K = data["cameras"].K
-<<<<<<< HEAD
 
             # Dirty but avoids creating error when running with real detector
             dt_det = 0
@@ -368,20 +341,6 @@ def get_predictions(self, pose_estimator: PoseEstimator) -> Dict[str, PoseEstima
                 sam_detections = make_detections_from_object_data(list_object_data).to(device)
                 sam_detections.infos['score'] = scores
 
-=======
-
-            # Temporary solution
-            if self.inference_cfg.detection_type == "sam":
-                list_object_data = []
-                scene_id = data['im_infos'][0]['scene_id']
-                view_id = data['im_infos'][0]['view_id']
-                print("scene and view :", scene_id, view_id)
-                list_object = object_data.loc[(object_data['scene_id'] == scene_id) & (object_data['image_id'] == view_id)].to_dict('records')
-                for object in list_object:
-                    list_object_data.append(ObjectData.from_json(object))
-                sam_detections = make_detections_from_object_data(list_object_data).to(device)
-                print("sam_detections =", sam_detections)
->>>>>>> c6cd60e ( first attempt at bop challenge)
             else:
                 sam_detections = None
             gt_detections = data["gt_detections"].cuda()
@@ -417,9 +376,9 @@ def get_predictions(self, pose_estimator: PoseEstimator) -> Dict[str, PoseEstima
                 v.infos['view_id'] = view_id
                 predictions_list[k].append(v)
 
-        # Concatenate the lists of PandasTensorCollections
+    # Concatenate the lists of PandasTensorCollections
         predictions = dict()
         for k, v in predictions_list.items():
             predictions[k] = tc.concatenate(v)
 
-        return predictions
+        return predictions

happypose/pose_estimators/megapose/src/megapose/inference/pose_estimator.py

@@ -582,12 +582,7 @@ def run_inference_pipeline(
                     detections, **detection_filter_kwargs
                 )
 
-<<<<<<< HEAD
             # Run the coarse estimator using detections
-=======
-            # Run the coarse estimator using gt_detections
-            print("detections coarse =", detections)
->>>>>>> c6cd60e ( first attempt at bop challenge)
             data_TCO_coarse, coarse_extra_data = self.forward_coarse_model(
                 observation=observation,
                 detections=detections,
@@ -679,4 +674,4 @@ def filter_pose_estimates(
 
         data_TCO_filtered = data_TCO[df.index.tolist()]
 
-        return data_TCO_filtered
+        return data_TCO_filtered

happypose/pose_estimators/megapose/src/megapose/scripts/run_full_megapose_eval.py

@@ -206,8 +206,6 @@ def update_cfg_debug(cfg: EvalConfig) -> FullEvalConfig:
     cli_cfg = OmegaConf.from_cli()
     logger.info(f"CLI config: \n {OmegaConf.to_yaml(cli_cfg)}")
 
-    print("full eval config =", FullEvalConfig.ds_names)
-
     cfg: FullEvalConfig = OmegaConf.structured(FullEvalConfig)
     cfg.hardware = HardwareConfig(
         n_cpus=int(os.environ.get("N_CPUS", 10)),