A very rough skeleton

ethology/annotations/__init__.py

@@ -0,0 +1 @@
+"""Utils to validate, fix, combine annotations, etc."""

ethology/datasets/__init__.py

@@ -0,0 +1 @@
+""" Utils to define datasets for training. """

ethology/detectors/__init__.py

@@ -0,0 +1,12 @@
+"""Detectors, pose estimation, detector+tracking models"""
+
+# If we use Pytorch lightning, here we will have LightningModules for each model
+
+# Pytorch lightning
+# - all definitions of what will happen in training are in a LightningModule 
+# - The [lightning module](https://pytorch-lightning.readthedocs.io/en/1.5.10/common/lightning_module.html) holds all the core research ingredients:
+# 		- The model
+# 		- The optimizers
+# 		- The train/ val/ test steps
+
+# Maybe following ultralytics structure, with subdirectories model, predict, train, val

ethology/napari/__init__.py

@@ -0,0 +1 @@
+"""Visualisation using a napari GUI."""

ethology/stereo/__init__.py

@@ -0,0 +1 @@
+"""Utils to compute 3D positions from 2D data from cameras in stereo config."""

ethology/trackers/__init__.py

@@ -0,0 +1 @@
+"""Trackers to apply on bboxes or kpt detections."""

ethology/video/__init__.py

@@ -0,0 +1 @@
+"""Utils to extract frames, reencode videos, crop, etc. """

ethology/visualisation/__init__.py

@@ -0,0 +1 @@
+"""Utils to visualise results or generate plots."""