#!/bin/bash ###################################################################################### # set the parameters for whole scene reconstruction below # for defination of these parameters, please refer to the recon.py ###################################################################################### KEYFRAME_STRIDE=20 UPDATE_BUFFER_INTV=3 MAX_NUM_REGISTER=10 WIN_R=5 NUM_SCENE_FRAME=10 INITIAL_WINSIZE=5 CONF_THRES_I2P=1.5 # the parameter below have nothing to do with the evaluation NUM_POINTS_SAVE=1000000 CONF_THRES_L2W=10 GPU_ID=-1 SCENE_NAMES=("office0" "office1" "office2" "office3" "office4" "room0" "room1" "room2") for SCENE_NAME in ${SCENE_NAMES[@]}; do TEST_NAME="Replica_${SCENE_NAME}" echo "--------Start reconstructing scene ${SCENE_NAME} with test name ${TEST_NAME}--------" python recon.py \ --test_name "${TEST_NAME}" \ --img_dir "data/Replica/${SCENE_NAME}/results" \ --gpu_id $GPU_ID \ --keyframe_stride $KEYFRAME_STRIDE \ --win_r $WIN_R \ --num_scene_frame $NUM_SCENE_FRAME \ --initial_winsize $INITIAL_WINSIZE \ --conf_thres_l2w $CONF_THRES_L2W \ --conf_thres_i2p $CONF_THRES_I2P \ --num_points_save $NUM_POINTS_SAVE \ --update_buffer_intv $UPDATE_BUFFER_INTV \ --max_num_register $MAX_NUM_REGISTER \ --save_for_eval echo "--------Start evaluating scene ${SCENE_NAME} with test name ${TEST_NAME}--------" python evaluation/eval_recon.py \ --test_name="${TEST_NAME}" \ --gt_pcd="results/gt/replica/${SCENE_NAME}_pcds.npy" done