feat: flexible numpy/torch backend and tensor type checking (#69)

This PR is a combination of:
* feat: automatic tensor backend and type checks (#64)
* perf: performance improvements for backend function wrappers (#66)
* feat: improved backend, union handling, tensor creation APIs (#68)

.github/workflows/unit_test.yml

@@ -25,6 +25,12 @@ jobs:
         run: |
           python -m pip install --upgrade pip
           pip install -e .[dev]
-      - name: Run unit tests
+      - name: Run unit tests (numpy)
+        run: |
+          pytest
+      - name: Install dependencies with torch
+        run: |
+          pip install -e .[torch]
+      - name: Run unit tests (numpy + torch)
         run: |
           pytest

.gitignore

@@ -1,6 +1,9 @@
 camtools/assets/bbox_box.png
 camtools/assets/bbox_box_blender.png
 
+# cProfile
+*.prof
+
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]

camtools/__init__.py

@@ -1,4 +1,5 @@
 from . import artifact
+from . import backend
 from . import camera
 from . import colmap
 from . import colormap
@@ -16,7 +17,6 @@
 from . import transform
 from . import util
 
-
 try:
     # Python >= 3.8
     from importlib.metadata import version