readme.md

README.md

@@ -3,62 +3,142 @@
 
 YOLOv4, YOLOv4-tiny Implemented in Tensorflow 2.0. 
 Convert YOLO v4, YOLOv3, YOLO tiny .weights to .pb, .tflite and trt format for tensorflow, tensorflow lite, tensorRT.
+<p align="center"><img src="data/helpers/demo.gif"\></p>
 
-Download yolov4.weights file: https://drive.google.com/open?id=1cewMfusmPjYWbrnuJRuKhPMwRe_b9PaT
+## Getting Started
+### Conda (Recommended)
 
+```bash
+# Tensorflow CPU
+conda env create -f conda-cpu.yml
+conda activate yolov4-cpu
+
+# Tensorflow GPU
+conda env create -f conda-gpu.yml
+conda activate yolov4-gpu
+```
 
-### Prerequisites
-* Tensorflow 2.3.0rc0
+### Pip
+```bash
+# TensorFlow CPU
+pip install -r requirements.txt
+
+# TensorFlow GPU
+pip install -r requirements-gpu.txt
+```
+### Nvidia Driver (For GPU, if you are not using Conda Environment and haven't set up CUDA yet)
+Make sure to use CUDA Toolkit version 10.1 as it is the proper version for the TensorFlow version used in this repository.
+https://developer.nvidia.com/cuda-10.1-download-archive-update2
 
 ### Performance
-<p align="center"><img src="data/performance.png" width="640"\></p>
+Check out how YOLOv4 compares to other object detection systems.
+
+<p align="center"><img src="data/helpers/performance.png" width="640"\></p>
+
+## Downloading Official Pre-trained Weights
+YOLOv4 comes pre-trained and able to detect 80 classes. For easy demo purposes we will use the pre-trained weights.
+Download pre-trained yolov4.weights file: https://drive.google.com/open?id=1cewMfusmPjYWbrnuJRuKhPMwRe_b9PaT
+
+Copy and paste yolov4.weights from your downloads folder into the 'data' folder of this repository.
+
+If you want to use yolov4-tiny.weights, a smaller model that is faster at running detections but less accurate, download file here: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.weights
+
+## Using Custom Trained YOLOv4 Weights
+<strong>Learn How To Train Custom YOLOv4 Weights here: https://www.youtube.com/watch?v=mmj3nxGT2YQ </strong>
+
+USE MY LICENSE PLATE TRAINED CUSTOM WEIGHTS: https://drive.google.com/file/d/1EUPtbtdF0bjRtNjGv436vDY28EN5DXDH/view?usp=sharing
+
+Copy and paste your custom .weights file into the 'data' folder and copy and paste your custom .names into the 'data/classes/' folder.
+
+The only change within the code you need to make in order for your custom model to work is on line 14 of 'core/config.py' file.
+Update the code to point at your custom .names file as seen below. (my custom .names file is called custom.names but yours might be named differently)
+<p align="center"><img src="data/helpers/custom_config.png" width="640"\></p>
 
-### Demo
+<strong>Note:</strong> If you are using the pre-trained yolov4 then make sure that line 14 remains <strong>coco.names</strong>.
 
+## YOLOv4 Using Tensorflow (tf, .pb model)
+To implement YOLOv4 using TensorFlow, first we convert the .weights into the corresponding TensorFlow model files and then run the model.
 ```bash
 # Convert darknet weights to tensorflow
 ## yolov4
 python save_model.py --weights ./data/yolov4.weights --output ./checkpoints/yolov4-416 --input_size 416 --model yolov4 
 
-## yolov4-tiny
+# yolov4-tiny
 python save_model.py --weights ./data/yolov4-tiny.weights --output ./checkpoints/yolov4-tiny-416 --input_size 416 --model yolov4 --tiny
 
-# Run demo tensorflow
-python detect.py --weights ./checkpoints/yolov4-416 --size 416 --model yolov4 --image ./data/kite.jpg
+# custom yolov4
+python save_model.py --weights ./data/custom.weights --output ./checkpoints/custom-416 --input_size 416 --model yolov4 
 
-python detect.py --weights ./checkpoints/yolov4-tiny-416 --size 416 --model yolov4 --image ./data/kite.jpg --tiny
+# Run yolov4 tensorflow model
+python detect.py --weights ./checkpoints/yolov4-416 --size 416 --model yolov4 --images ./data/images/kite.jpg
 
+# Run yolov4-tiny tensorflow model
+python detect.py --weights ./checkpoints/yolov4-tiny-416 --size 416 --model yolov4 --images ./data/images/kite.jpg --tiny
+
+# Run custom yolov4 tensorflow model
+python detect.py --weights ./checkpoints/custom-416 --size 416 --model yolov4 --images ./data/images/car.jpg
+
+# Run yolov4 on video
+python detect_video.py --weights ./checkpoints/yolov4-416 --size 416 --model yolov4 --video ./data/video/video.mp4 --output ./detections/results.avi
+
+# Run custom yolov4 model on video
+python detect_video.py --weights ./checkpoints/custom-416 --size 416 --model yolov4 --video ./data/video/cars.mp4 --output ./detections/results.avi
+
+# Run yolov4 on webcam
+python detect_video.py --weights ./checkpoints/yolov4-416 --size 416 --model yolov4 --video 0 --output ./detections/results.avi
 ```
-If you want to run yolov3 or yolov3-tiny change ``--model yolov3`` in command
+If you want to run yolov3 or yolov3-tiny change ``--model yolov3`` and .weights file in above commands.
 
-#### Output
+<strong>Note:</strong> You can also run the detector on multiple images at once by changing the --images flag like such ``--images "./data/images/kite.jpg, ./data/images/dog.jpg"``
 
-##### Yolov4 original weight
-<p align="center"><img src="result.png" width="640"\></p>
+### Result Image(s) (Regular TensorFlow)
+You can find the outputted image(s) showing the detections saved within the 'detections' folder.
+#### Pre-trained YOLOv4 Model Example
+<p align="center"><img src="data/helpers/result.png" width="640"\></p>
 
-##### Yolov4 tflite int8
-<p align="center"><img src="result-int8.png" width="640"\></p>
+#### Custom YOLOv4 Model Example (see video link above to train this model)
+<p align="center"><img src="data/helpers/custom_result.png" width="640"\></p>
 
-### Convert to tflite
+### Result Video
+Video saves wherever you point --output flag to. If you don't set the flag then your video will not be saved with detections on it.
+<p align="center"><img src="data/helpers/demo.gif"\></p>
 
+## YOLOv4 Using TensorFlow Lite (.tflite model)
+Can also implement YOLOv4 using TensorFlow Lite. TensorFlow Lite is a much smaller model and perfect for mobile or edge devices (raspberry pi, etc).
 ```bash
 # Save tf model for tflite converting
 python save_model.py --weights ./data/yolov4.weights --output ./checkpoints/yolov4-416 --input_size 416 --model yolov4 --framework tflite
 
+# Save custom yolov4 tf model for tflite converting
+python save_model.py --weights ./data/custom.weights --output ./checkpoints/custom-416 --input_size 416 --model yolov4 --framework tflite
+
 # yolov4
 python convert_tflite.py --weights ./checkpoints/yolov4-416 --output ./checkpoints/yolov4-416.tflite
 
+# convert custom yolov4 tflite model
+python convert_tflite.py --weights ./checkpoints/custom-416 --output ./checkpoints/custom-416.tflite
+
 # yolov4 quantize float16
 python convert_tflite.py --weights ./checkpoints/yolov4-416 --output ./checkpoints/yolov4-416-fp16.tflite --quantize_mode float16
 
 # yolov4 quantize int8
 python convert_tflite.py --weights ./checkpoints/yolov4-416 --output ./checkpoints/yolov4-416-int8.tflite --quantize_mode int8 --dataset ./coco_dataset/coco/val207.txt
 
-# Run demo tflite model
-python detect.py --weights ./checkpoints/yolov4-416.tflite --size 416 --model yolov4 --image ./data/kite.jpg --framework tflite
+# Run tflite model
+python detect.py --weights ./checkpoints/yolov4-416.tflite --size 416 --model yolov4 --images ./data/images/kite.jpg --framework tflite
+
+# Run custom tflite model
+python detect.py --weights ./checkpoints/custom-416.tflite --size 416 --model yolov4 --images ./data/images/car.jpg --framework tflite
 ```
+### Result Image (TensorFlow Lite)
+You can find the outputted image(s) showing the detections saved within the 'detections' folder.
+#### TensorFlow Lite int8 Example
+<p align="center"><img src="data/helpers/result-int8.png" width="640"\></p>
+
 Yolov4 and Yolov4-tiny int8 quantization have some issues. I will try to fix that. You can try Yolov3 and Yolov3-tiny int8 quantization 
-### Convert to TensorRT
+
+## YOLOv4 Using TensorRT
+Can also implement YOLOv4 using TensorFlow's TensorRT. TensorRT is a high-performance inference optimizer and runtime that can be used to perform inference in lower precision (FP16 and INT8) on GPUs. TensorRT can allow up to 8x higher performance than regular TensorFlow.
 ```bash# yolov3
 python save_model.py --weights ./data/yolov3.weights --output ./checkpoints/yolov3.tf --input_size 416 --model yolov3
 python convert_trt.py --weights ./checkpoints/yolov3.tf --quantize_mode float16 --output ./checkpoints/yolov3-trt-fp16-416
@@ -70,9 +150,70 @@ python convert_trt.py --weights ./checkpoints/yolov3-tiny.tf --quantize_mode flo
 # yolov4
 python save_model.py --weights ./data/yolov4.weights --output ./checkpoints/yolov4.tf --input_size 416 --model yolov4
 python convert_trt.py --weights ./checkpoints/yolov4.tf --quantize_mode float16 --output ./checkpoints/yolov4-trt-fp16-416
+python detect.py --weights ./checkpoints/yolov4-trt-fp16-416 --model yolov4 --images ./data/images/kite.jpg --framework trt
 ```
 
-### Evaluate on COCO 2017 Dataset
+## Command Line Args Reference
+
+```bash
+save_model.py:
+  --weights: path to weights file
+    (default: './data/yolov4.weights')
+  --output: path to output
+    (default: './checkpoints/yolov4-416')
+  --[no]tiny: yolov4 or yolov4-tiny
+    (default: 'False')
+  --input_size: define input size of export model
+    (default: 416)
+  --framework: what framework to use (tf, trt, tflite)
+    (default: tf)
+  --model: yolov3 or yolov4
+    (default: yolov4)
+
+detect.py:
+  --images: path to input images as a string with images separated by ","
+    (default: './data/images/kite.jpg')
+  --output: path to output folder
+    (default: './detections/')
+  --[no]tiny: yolov4 or yolov4-tiny
+    (default: 'False')
+  --weights: path to weights file
+    (default: './checkpoints/yolov4-416')
+  --framework: what framework to use (tf, trt, tflite)
+    (default: tf)
+  --model: yolov3 or yolov4
+    (default: yolov4)
+  --size: resize images to
+    (default: 416)
+  --iou: iou threshold
+    (default: 0.45)
+  --score: confidence threshold
+    (default: 0.25)
+
+detect_video.py:
+  --video: path to input video (use 0 for webcam)
+    (default: './data/video/video.mp4')
+  --output: path to output video (remember to set right codec for given format. e.g. XVID for .avi)
+    (default: None)
+  --output_format: codec used in VideoWriter when saving video to file
+    (default: 'XVID)
+  --[no]tiny: yolov4 or yolov4-tiny
+    (default: 'false')
+  --weights: path to weights file
+    (default: './checkpoints/yolov4-416')
+  --framework: what framework to use (tf, trt, tflite)
+    (default: tf)
+  --model: yolov3 or yolov4
+    (default: yolov4)
+  --size: resize images to
+    (default: 416)
+  --iou: iou threshold
+    (default: 0.45)
+  --score: confidence threshold
+    (default: 0.25)
+```
+
+## Evaluate on COCO 2017 Dataset
 ```bash
 # run script in /script/get_coco_dataset_2017.sh to download COCO 2017 Dataset
 # preprocess coco dataset
@@ -98,7 +239,7 @@ python main.py --output results_yolov4_tf
 | YoloV3      | 55.43   | 52.32   |         |
 | YoloV4      | 61.96   | 57.33   |         |
 
-### Benchmark
+## Benchmark
 ```bash
 python benchmarks.py --size 416 --model yolov4 --weights ./data/yolov4.weights
 ```
@@ -144,7 +285,7 @@ python benchmarks.py --size 416 --model yolov4 --weights ./data/yolov4.weights
 | YoloV3 FPS  |         |         |         |
 | YoloV4 FPS  |         |         |         |
 
-### Traning your own model
+## Traning your own model in TensorFlow
 ```bash
 # Prepare your dataset
 # If you want to train from scratch:
@@ -157,7 +298,7 @@ python train.py --weights ./data/yolov4.weights
 ```
 The training performance is not fully reproduced yet, so I recommended to use Alex's [Darknet](https://github.com/AlexeyAB/darknet) to train your own data, then convert the .weights to tensorflow or tflite.
 
-
+<strong>Use this video to train your own model easily in Google Colab: https://www.youtube.com/watch?v=mmj3nxGT2YQ </strong>
 
 ### TODO
 * [x] Convert YOLOv4 to TensorRT

benchmarks.py

@@ -15,7 +15,7 @@
 flags.DEFINE_string('framework', 'tf', '(tf, tflite, trt')
 flags.DEFINE_string('model', 'yolov4', 'yolov3 or yolov4')
 flags.DEFINE_string('weights', './data/yolov4.weights', 'path to weights file')
-flags.DEFINE_string('image', './data/kite.jpg', 'path to input image')
+flags.DEFINE_string('images', './data/images/kite.jpg', 'path to input image')
 flags.DEFINE_integer('size', 416, 'resize images to')
 
 

conda-cpu.yml

@@ -0,0 +1,15 @@
+name: yolov4-cpu
+
+dependencies:
+  - python==3.7
+  - pip
+  - matplotlib
+  - opencv
+  - pip:
+    - opencv-python==4.1.1.26
+    - lxml
+    - tqdm
+    - tensorflow==2.3.0rc0
+    - absl-py
+    - easydict
+    - pillow

conda-gpu.yml

@@ -0,0 +1,17 @@
+name: yolov4-gpu
+
+dependencies:
+  - python==3.7
+  - pip
+  - matplotlib
+  - opencv
+  - cudnn
+  - cudatoolkit==10.1.243
+  - pip:
+    - tensorflow-gpu==2.3.0rc0
+    - opencv-python==4.1.1.26
+    - lxml
+    - tqdm
+    - absl-py
+    - easydict
+    - pillow

core/utils.py

@@ -124,7 +124,7 @@ def image_preprocess(image, target_size, gt_boxes=None):
         gt_boxes[:, [1, 3]] = gt_boxes[:, [1, 3]] * scale + dh
         return image_paded, gt_boxes
 
-def draw_bbox(image, bboxes, classes=read_class_names(cfg.YOLO.CLASSES), show_label=True):
+def draw_bbox(image, bboxes, classes=read_class_names(cfg.YOLO.CLASSES), allowed_classes=list(read_class_names(cfg.YOLO.CLASSES).values()), show_label=True):
     num_classes = len(classes)
     image_h, image_w, _ = image.shape
     hsv_tuples = [(1.0 * x / num_classes, 1., 1.) for x in range(num_classes)]
@@ -147,19 +147,25 @@ def draw_bbox(image, bboxes, classes=read_class_names(cfg.YOLO.CLASSES), show_la
         fontScale = 0.5
         score = out_scores[0][i]
         class_ind = int(out_classes[0][i])
-        bbox_color = colors[class_ind]
-        bbox_thick = int(0.6 * (image_h + image_w) / 600)
-        c1, c2 = (coor[1], coor[0]), (coor[3], coor[2])
-        cv2.rectangle(image, c1, c2, bbox_color, bbox_thick)
-
-        if show_label:
-            bbox_mess = '%s: %.2f' % (classes[class_ind], score)
-            t_size = cv2.getTextSize(bbox_mess, 0, fontScale, thickness=bbox_thick // 2)[0]
-            c3 = (c1[0] + t_size[0], c1[1] - t_size[1] - 3)
-            cv2.rectangle(image, c1, (np.float32(c3[0]), np.float32(c3[1])), bbox_color, -1) #filled
-
-            cv2.putText(image, bbox_mess, (c1[0], np.float32(c1[1] - 2)), cv2.FONT_HERSHEY_SIMPLEX,
-                        fontScale, (0, 0, 0), bbox_thick // 2, lineType=cv2.LINE_AA)
+        class_name = classes[class_ind]
+
+        # check if class is in allowed classes
+        if class_name not in allowed_classes:
+            continue
+        else:
+            bbox_color = colors[class_ind]
+            bbox_thick = int(0.6 * (image_h + image_w) / 600)
+            c1, c2 = (coor[1], coor[0]), (coor[3], coor[2])
+            cv2.rectangle(image, c1, c2, bbox_color, bbox_thick)
+
+            if show_label:
+                bbox_mess = '%s: %.2f' % (classes[class_ind], score)
+                t_size = cv2.getTextSize(bbox_mess, 0, fontScale, thickness=bbox_thick // 2)[0]
+                c3 = (c1[0] + t_size[0], c1[1] - t_size[1] - 3)
+                cv2.rectangle(image, c1, (np.float32(c3[0]), np.float32(c3[1])), bbox_color, -1) #filled
+
+                cv2.putText(image, bbox_mess, (c1[0], np.float32(c1[1] - 2)), cv2.FONT_HERSHEY_SIMPLEX,
+                            fontScale, (0, 0, 0), bbox_thick // 2, lineType=cv2.LINE_AA)
     return image
 
 def bbox_iou(bboxes1, bboxes2):