Fix CI

.github/workflows/main.yml

@@ -1,21 +1,24 @@
-name: Ubuntu CI
+name: CI
 
-on: [push, pull_request, workflow_dispatch]
+on: [pull_request, workflow_dispatch]
 
 jobs:
   main:
-    runs-on: ubuntu-18.04
+    runs-on: ${{ matrix.os }}
+    strategy:
+        matrix:
+            os: [ubuntu-22.04, ubuntu-20.04, windows-latest]
     steps:
       - uses: actions/checkout@v2
 
       - name: Set up Python
         uses: actions/setup-python@v1
         with:
-          python-version: 3.7
+          python-version: 3.8
 
       - name: Upgrade pip
         run: python3 -m pip install --upgrade pip
-
+      
       - name: Install Poetry
         run: pip3 install poetry --user
 
@@ -25,21 +28,21 @@ jobs:
       # Prints the help pages of all scripts to see if the imports etc. work
       - name: Test the help pages
         run: |
-          poetry run yolo-train -h
-          poetry run yolo-test -h
-          poetry run yolo-detect -h
+          poetry run yoeo-train -h
+          poetry run yoeo-test -h
+          poetry run yoeo-detect -h
 
       - name: Demo Training
-        run: poetry run yolo-train --data config/custom.data  --model config/yolov3.cfg --epochs 30
+        run: poetry run yoeo-train --data config/custom.data --epochs 30
 
       - name: Demo Evaluate
-        run: poetry run yolo-test --data config/custom.data  --model config/yolov3.cfg --weights checkpoints/yolov3_ckpt_29.pth
+        run: poetry run yoeo-test --data config/custom.data --weights checkpoints/yoeo_checkpoint_29.pth
 
       - name: Demo Detect
-        run: poetry run yolo-detect --batch_size 2 --weights checkpoints/yolov3_ckpt_29.pth
+        run: poetry run yoeo-detect --batch_size 2 --weights checkpoints/yoeo_checkpoint_29.pth
 
   linter:
-    runs-on: ubuntu-18.04
+    runs-on: ubuntu-22.04
     steps:
       - uses: actions/checkout@v2
 

README.md

@@ -68,7 +68,7 @@ You can adjust the log directory using `--logdir <path>` when running `tensorboa
 ## Train on Custom Dataset
 
 #### Classes
-Add class names to `data/custom/classes.names`. This file should have one row per class name.
+Add class names to `data/custom/yoeo_names.yaml`.
 
 #### Image Folder
 Move the images of your dataset to `data/custom/images/`.
@@ -77,7 +77,7 @@ Move the images of your dataset to `data/custom/images/`.
 Move your yolo annotations to `data/custom/labels/`. The dataloader expects that the annotation file corresponding to the image `data/custom/images/train.jpg` has the path `data/custom/labels/train.txt`. Each row in the annotation file should define one bounding box, using the syntax `label_idx x_center y_center width height`. The coordinates should be scaled `[0, 1]`, and the `label_idx` should be zero-indexed and correspond to the row number of the class name in `data/custom/classes.names`.
 
 #### Segmentation Annotation Folder
-Move your segmentation annotations to `data/custom/segmentations/`. The dataloader expects that the annotation file corresponding to the image `data/custom/images/train.jpg` has the path `data/custom/segmentations/train.png`. The classes for each pixel are encoded via the class id.
+Move your segmentation annotations to `data/custom/yoeo_segmentations/`. The dataloader expects that the annotation file corresponding to the image `data/custom/images/train.jpg` has the path `data/custom/yoeo_segmentations/train.png`. The classes for each pixel are encoded via the class id.
 
 #### Define Train and Validation Sets
 In `data/custom/train.txt` and `data/custom/valid.txt`, add paths to images that will be used as train and validation data respectively.

config/custom.data

@@ -1,4 +1,3 @@
-classes= 1
 train=data/custom/train.txt
 valid=data/custom/valid.txt
-names=data/custom/class_names.yaml
+names=data/custom/yoeo_names.yaml

data/custom/labels/1028-img00671.txt

@@ -0,0 +1,4 @@
+2 0.7056451612903226 0.3854166666666667 0.10806451612903226 0.26666666666666666
+2 0.09032258064516129 0.4510416666666667 0.18064516129032257 0.47708333333333336
+1 0.3596774193548387 0.2677083333333333 0.07096774193548387 0.5354166666666667
+0 0.6508064516129032 0.475 0.0467741935483871 0.0625

data/custom/labels/138-16_02_2018__12_01_43_0417_upper.txt

@@ -0,0 +1,7 @@
+2 0.05625 0.19270833333333334 0.1125 0.08958333333333333
+2 0.1375 0.09583333333333334 0.075 0.12916666666666668
+2 0.228125 0.14479166666666668 0.090625 0.17708333333333334
+2 0.41953125 0.128125 0.1046875 0.22708333333333333
+2 0.65078125 0.12916666666666668 0.0671875 0.22916666666666666
+1 0.23671875 0.042708333333333334 0.0234375 0.08541666666666667
+1 0.00703125 0.11666666666666667 0.0140625 0.075

data/custom/train.txt

@@ -0,0 +1,2 @@
+data/custom/images/1028-img00671.png
+data/custom/images/138-16_02_2018__12_01_43_0417_upper.png

data/custom/valid.txt

@@ -0,0 +1,2 @@
+data/custom/images/1028-img00671.png
+data/custom/images/138-16_02_2018__12_01_43_0417_upper.png

data/custom/yoeo_names.yaml

@@ -0,0 +1,8 @@
+detection:
+    - ball
+    - goalpost
+    - robot
+segmentation:
+    - background
+    - lines
+    - field

yoeo/utils/loss.py

@@ -153,6 +153,16 @@ def build_targets(p, targets, model):
         # Check if we have targets
         if nt:
             t = t.squeeze(0)
+            # Check if we use anchor boxes
+            if na != 1:
+                # Calculate ratio between anchor and target box for both width and height
+                r = t[:, :, 4:6] / anchors[:, None]
+                # Select the ratios that have the highest divergence in any axis and check if the ratio is less than 4
+                j = torch.max(r, 1. / r).max(2)[0] < 4
+                # Only use targets that have the correct ratios for their anchors
+                # That means we only keep ones that have a matching anchor and we loose the anchor dimension
+                # The anchor id is still saved in the 7th value of each target
+                t = t[j]
         else:
             t = targets[0]