Evaluate script: review of command-line arguments (#172)

* move eval params to config (WIP)

* follow train CLI: add debugger options, add experiment name, use score_threshold from config

* fix prettier

* edit CLI defaults and get dataset params from ckpt if not defined (WIP)

* fix ninja comma

* Add sections to config

* Rename to evaluate utils

* Match current train script and add slurm logs as artifacts

* Fix evaluate_utils

* Use config from ckpt if not passed. Use dataset, annot files and seed from ckpt if not passed.

* Clarify CLI help (hopefully)

* Add score threshold for visualisation as CLI argument

* Small fix to config yaml

* Clean up

* Fix save frames and add output_dir

* Fix tests

* Move get_ functions to evaluate utils

* Replace assert by try-except

crabs/detection_tracking/config/faster_rcnn.yaml

@@ -1,12 +1,22 @@
+# Dataset
+#-------------
+train_fraction: 0.8
+val_over_test_fraction: 0.5
+num_workers: 4
+
+# -------------------
+# Model architecture
+# -------------------
+num_classes: 2
+
+# -------------------------------
+# Training & validation parameters
+# -------------------------------
 n_epochs: 250
 learning_rate: 0.00005
 wdecay: 0.00005
 batch_size_train: 4
-batch_size_test: 4
 batch_size_val: 4
-num_classes: 2
-train_fraction: 0.8
-val_over_test_fraction: 0.5
 checkpoint_saving:
   every_n_epochs: 50
   keep_last_n_ckpts: 5
@@ -17,8 +27,16 @@ checkpoint_saving:
   # if all, all checkpoints for every epoch are added as artifacts during training,
   # if False, no checkpoints are added as artifacts.
   save_last: True
+
+# -----------------------
+# Evaluation parameters
+# -----------------------
 iou_threshold: 0.1
-num_workers: 4
+batch_size_test: 4
+
+# -------------------
+# Data augmentation
+# -------------------
 transform_brightness: 0.5
 transform_hue: 0.3
 gaussian_blur_params:
@@ -28,6 +46,10 @@ gaussian_blur_params:
   sigma:
     - 0.1
     - 5.0
+
+# ----------------------------
+# Hyperparameter optimisation
+# -----------------------------
 # when we run Optuna, the n_trials and n_epochs above will be overwritten by the parameters set by Optuna
 optuna:
   # Parameters for hyperparameter optimisation with Optuna: