Skip to content

tests.md

Latest commit

 

History

History
107 lines (92 loc) · 3.99 KB

tests.md

File metadata and controls

107 lines (92 loc) · 3.99 KB
Raw

Usage

For any test, if it's unittest class, you can run

  • python -m unittest tests.xxx.test_file.py or
  • python -m unittest discover tests.xxx.test_dir

to auto run all unittest for checking.

Test a single func, use

  • python -m unittest tests.xxx.test_file.TestDict.tests_method.

Some tests do not use check but give visual results in tests/xxx/results

We set default cfg file path at __init__.py. You can change it and tests will read the config from it.

Many of the tests on geometry or model will save visible results in the result dir.

tests_common

Tests for common class. Directly obtained from common_trainer project.

tests_arcnerf

Tests for all nerf/3d related functions.

tests_datasets

Tests for datasets. Including showing cameras, avg cams and test cam trajectory.

tests_any_datasets

It will read datasets from default.yaml, save related visual results into results folder. Change the config file in tests_datasets.__init__ to other dataset cfg in dataset_dir.

tests_geometry

Test geometry function.

tests_mesh

Test mesh function, like cal normals, getting face centers.

tests_point_cloud

Test mesh function, like point cloud io.

tests_sphere

Test the sphere representation and circle/spiral line creation

tests_poses

Test the poses creation and modification functions.

tests_projection

Test projection functions. Make pixels and transform them to points, then project points back to pixels. Check the difference.

tests_transformation

Test function in transform, like normalization, rotation

tests_triangle

Test function in triangle, like getting norm, getting circumcircle, etc.

tests_ray

Test function in ray, like getting sphere-ray-intersection, point-ray distance, etc.

tests_volume

Test function in volume, like volume pts/line/faces generation and visualization, ray-volume intersection with coarse structure, etc.

tests_render

tests_camera

Test camera creation, get_rays and project rays into pixel, visualize ray and ray points Also check when we rescale camera and image, whether rays get projected correctly.

tests_ray_helper

Test func for ray sampling, resample cdf/pdf, etc.

tests_models

Test all the models and components for correctness and time.

tests_base_modules

tests_activation

Test activation functions.

tests_encoding

Test all kinds of encoder. Some encoders have different backend(torch/cuda/tcnn), will check correctness.

tests_encoding_benchmark

Test encoders speed and time.

tests_linear

Test linear layers

tests_linear_network

Test implicit/radiance function with encoder and linear network.

tests_tcnn_fusedmlp_network

Test implicit/radiance function with encoder and tcnn_fusemlp network. Much faster that torch linear.

tests_benchmark

Test all the model and get the speed/time.

tests_fg_model

Test fg model for ray sampling and forward.

tests_bkg_model

Test each bkg model like nerf++.

tests_nerf

Test NeRF model.

tests_mipnerf

Test MipNeRF model.

tests_nerfpp

Test NeRF model with nerf++ as bkg model.

tests_neus

Test Neus model and its sdf_to_alpha method with sampling.

tests_volsdf

Test VolSDF model and its sdf_to_sigma method with sampling.

tests_ops

Tests of the customized operations in CUDA extension.

tests_loss

Tests for loss functions. Including geo_loss, img_loss, mask_loss.

tests_metric

Tests for eval metric functions. Including img_metric