Fork of SimpleRecon that includes a very simple test script. See here for the full README. See below for instructions on how to run test_simple.py in this repo.
Assuming a fresh Anaconda distribution, you can install dependencies with:
conda env create -f simplerecon_env.ymlDownload a pretrained model into the weights/ folder.
We provide the following models (scores are with online default keyframes):
--config |
Model | Abs Diff↓ | Sq Rel↓ | delta < 1.05↑ | Chamfer↓ | F-Score↑ |
|---|---|---|---|---|---|---|
hero_model.yaml |
Metadata + Resnet Matching | 0.0868 | 0.0127 | 74.26 | 5.69 | 0.680 |
dot_product_model.yaml |
Dot Product + Resnet Matching | 0.0910 | 0.0134 | 71.90 | 5.92 | 0.667 |
hero_model is the main model.
Run the following (note: most of the arguments are ignored). This will ask you to select the folder with images; navigate to the folder and double click on it; then select "Open". This assumes you've already run collect_images_for_fusion.py from here.
CUDA_VISIBLE_DEVICES=0 python test_simple.py --name HERO_MODEL --output_base_path OUTPUT_PATH --config_file configs/models/hero_model.yaml --load_weights_from_checkpoint weights/hero_model.ckpt --data_config configs/data/vdr_dense.yaml --num_workers 8 --batch_size 1 --fast_cost_volume --run_fusion --depth_fuser open3d --fuse_color \
This will save a depth image (in millimeters) in the simplerecon root folder.
If you want to visualize the saved depth image, you can run the following (e.g., with ipython):
import numpy as np
import matplotlib.pyplot as plt
depth = np.load("./depth.millimeters.frame-000007.npy")
plt.imshow(depth); plt.show()