构建no-text数据集

build_text_dataset.sh

@@ -0,0 +1,29 @@
+SEG_ROOT="/home/cyx6123/DuodenoVLA/data/ACT/aloha/act/data/raw_data/01-cannulation"
+OUT_DIR="/home/cyx6123/DuodenoVLA/data/ACT/aloha/act/data/cannulation"
+SCRIPT="/home/cyx6123/DuodenoVLA/data/ACT/aloha/act/build_endoscope_act_dataset.py"
+
+mkdir -p "$OUT_DIR"
+i=12
+for d in "$SEG_ROOT"/seg_*; do
+  [ -d "$d" ] || continue
+  echo "Building $d -> episode_$i"
+  python "$SCRIPT" \
+    --segment_dir "$d" \
+    --output_dir "$OUT_DIR" \
+    --episode_idx "$i" \
+    --max_frames -1 \
+    --camera_name top \
+    --crop 733 30 1754 1051 \
+    --resize 224 224 \
+    --instruction_template 'Cannulate the {label} on the phantom located at the {region} with the sphincterotome.' \
+    --instruction_empty 'No target visible.' \
+    --stop_instruction 'Stop move.' \
+    --motion_window 3 \
+    --motion_threshold 0.05 \
+    --state_norm minus1_1 \
+    --action_norm minus1_1 \
+    --encode_text_features \
+    --text_model_name distilbert-base-uncased \
+    --text_batch_size 32
+  i=$((i+1))
+done