-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathseg_scannet.sh
executable file
·30 lines (26 loc) · 1.2 KB
/
seg_scannet.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
#region growing on scannetpp
# source YOUR_CONDA_PATH/bin/activate opensai
#set the parameters below
VIEW_FREQ=5 # sample frequency of the frames
THRES_CONNECT="0.9,0.5,5" # dynamic threshold for region growing
MAX_NEIGHBOR_DISTANCE=2 # farthest distance to take neighbors into account
THRES_MERGE=200 # merge small groups with less than THRES_MERGE points during post-processing
DIS_DECAY=0.5 # decay rate of the distance weight
SIMILAR_METRIC="2-norm" # metric for similarity measurement
MASK_NAME="semantic-sam" # mask name for loading mask
ALIAS_MASK_NAME="semantic-sam" # mask name for saving results
TEXT_HEAD="demo_scannet"
TEXT="${TEXT_HEAD}_${VIEW_FREQ}view"
HEAD="${TEXT_HEAD}_${VIEW_FREQ}view_merge${THRES_MERGE}_${SIMILAR_METRIC}_${ALIAS_MASK_NAME}_connect(${THRES_CONNECT})_depth${MAX_NEIGHBOR_DISTANCE}"
EVAL_DIR="data/ScanNet/results/${HEAD}" # directory to export results
python sai3d.py \
--thres_merge=$THRES_MERGE \
--similar_metric=$SIMILAR_METRIC \
--thres_connect=$THRES_CONNECT \
--mask_name=$MASK_NAME \
--text=$TEXT \
--max_neighbor_distance=$MAX_NEIGHBOR_DISTANCE \
--view_freq=$VIEW_FREQ \
--use_torch \
--dis_decay=$DIS_DECAY \
--eval_dir=$EVAL_DIR \