add preprocessing script for TUM dataset

convenience_scripts/tum/taijing/tum_freiburg2_desk_preprocess.sh

@@ -0,0 +1,31 @@
+SLAMDIR=/home/tiejean/Documents/projects/ut_vslam
+DARADIR=/home/tiejean/Documents/mnt/oslam
+ORB_OUT=$DARADIR/orb_out
+ORB_POST_PROCESS=$DARADIR/orb_post_process
+
+bagname=freiburg2_desk
+configname=base7a_2_fallback_a_2
+
+configfile=$SLAMDIR/config/${base7a_2_fallback_a_2}.json
+calibration_dir=$DARADIR/calibration/
+orb_out_dir=$ORB_OUT/$bagname/
+unsparsified_orb_out=$ORB_POST_PROCESS/unsparsified_ut_vslam_in/$bagname/
+sparsified_orb_out=$ORB_POST_PROCESS/sparsified_ut_vslam_in/$configname/$bagname/
+
+mkdir -p $orb_out_dir
+mkdir -p $unsparsified_orb_out
+mkdir -p $sparsified_orb_out
+
+python3 src/data_preprocessing_utils/orb_stereo_reformat_data.py \
+    -i $orb_out_dir -o $unsparsified_orb_out
+
+./bin/initialize_traj_and_feats_from_orb_out \
+    --raw_data_path $orb_out_dir \
+    --calibration_path $calibration_dir \
+    --processed_data_path $unsparsified_orb_out
+
+./bin/orb_trajectory_sparsifier \
+    --param_prefix $bagname \
+    --input_processed_data_path $unsparsified_orb_out \
+    --output_processed_data_path $sparsified_orb_out \
+    --params_config_file $configfile

src/data_preprocessing_utils/orb_stereo_reformat_data.py

@@ -24,7 +24,7 @@ def add_to_features(features_dict, line):
             return
         measurement_x2 = float(tokens[4])
         measurement_y2 = float(tokens[5])
-    except ValueError as e:
+    except:
         return
     features_dict[feature_id] = (
         depth, measurement_x1, measurement_y1, measurement_x2, measurement_y2)

src/evaluation/objects/convert_object_detections_from_json_to_csv.py

@@ -0,0 +1,76 @@
+import json
+import argparse
+import os
+
+# https://github.com/ultralytics/yolov5/blob/master/data/coco.yaml
+kIds2Clsnames = {0:"person",
+                 14:"bird", 
+                 15:"cat", 
+                 16:"dog", 
+                 23:"giraffe", 
+                 26:"handbag", 
+                 32:"sports ball", 
+                 39:"bottle", 
+                 41:"cup", 
+                 42:"fork",
+                 43:"knife",
+                 44:"spoon",
+                 45:"bowl",
+                 47:"apple",
+                 51:"carrot",
+                 54:"donut",
+                 56:"chair",
+                 58:"potted plant",
+                 60:"dinning table",
+                 62:"tv",
+                 64:"mouse",
+                 65:"remote",
+                 66:"keyboard",
+                 67:"cell phone", 
+                 73:"book", 
+                 75: "vase",
+                 76: "scissors",
+                 77:"teddy bear", }
+kDelimiter = ", "
+
+def double2Ints(time: float):
+    return int(time), int((time-int(time))*1e9)
+
+def parse_args():
+    parser = argparse.ArgumentParser(description="Convert object detections from OA-SLAM to ObVi-SLAM format") 
+    parser.add_argument("--inpath", required=True, type=str, help="input detection file .json")
+    parser.add_argument("--outpath", required=True, type=str, help="output detection file .csv")
+    args = parser.parse_args()
+    return args
+
+if __name__ == "__main__":
+    args = parse_args()
+    ifp = open(args.inpath, "r")
+    if ifp.closed:
+        raise FileNotFoundError("Failed to open file " + args.inpath)
+    ofp = open(args.outpath, "w")
+    if ofp.closed:
+        raise FileNotFoundError("Failed to open file " + args.outpath)
+    data = json.load(ifp)
+    ofp.write("min_pixel_x, min_pixel_y, max_pixel_x, max_pixel_y, semantic_class, seconds, nano_seconds, camera_id, detection_confidence\n")
+
+    for fdata in data:
+        seconds, nano_seconds = double2Ints(float(fdata["file_name"][:-4]))
+        camera_id = 1
+        for detection in fdata["detections"]:
+            minx,miny,maxx,maxy = detection["bbox"][0],detection["bbox"][1],detection["bbox"][2],detection["bbox"][3]
+            cls = detection["category_id"]
+            if cls not in kIds2Clsnames.keys():
+                print(cls)
+                continue
+            cls = kIds2Clsnames[cls]
+            detection_confidence = detection["detection_score"]
+            line = [ minx,miny,maxx,maxy,cls,seconds,nano_seconds,camera_id,detection_confidence]
+            line = [str(token) for token in line]
+            ofp.write(line[0])
+            for token in line[1:]:
+                ofp.write(kDelimiter + token)
+            ofp.write("\n")
+
+    ifp.close()
+    ofp.close()