Upload 8 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+examples/3.jpg filter=lfs diff=lfs merge=lfs -text

Base-RCNN-FPN.yaml

@@ -0,0 +1,50 @@
+MODEL:
+    META_ARCHITECTURE: "GeneralizedRCNN"
+
+    BACKBONE:
+        NAME: "build_resnet_fpn_backbone"
+
+    RESNETS:
+        OUT_FEATURES: ["res2", "res3", "res4", "res5"]
+
+    FPN:
+        IN_FEATURES: ["res2", "res3", "res4", "res5"]
+
+    ANCHOR_GENERATOR:
+        # One size for each in feature map
+        SIZES: [[32], [64], [128], [256], [512]]
+        # Three aspect ratios (same for all in feature maps)
+        ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
+
+    RPN:
+        IN_FEATURES: ["p2", "p3", "p4", "p5", "p6"]
+        PRE_NMS_TOPK_TRAIN: 2000  # Per FPN level
+        PRE_NMS_TOPK_TEST: 1000  # Per FPN level
+        # Detectron1 uses 2000 proposals per-batch,
+        # (See "modeling/rpn/rpn_outputs.py" for details of this legacy issue)
+        # which is approximately 1000 proposals per-image since the default
+        # batch size for FPN is 2.
+        POST_NMS_TOPK_TRAIN: 1000
+        POST_NMS_TOPK_TEST: 1000
+
+    ROI_HEADS:
+        NAME: "StandardROIHeads"
+        IN_FEATURES: ["p2", "p3", "p4", "p5"]
+
+    ROI_BOX_HEAD:
+        NAME: "FastRCNNConvFCHead"
+        NUM_FC: 2
+        POOLER_RESOLUTION: 7
+
+    ROI_MASK_HEAD:
+        NAME: "MaskRCNNConvUpsampleHead"
+        NUM_CONV: 4
+        POOLER_RESOLUTION: 14
+
+INPUT:
+    MIN_SIZE_TRAIN: (640, 672, 704, 736, 768, 800)
+
+SOLVER:
+    CHECKPOINT_PERIOD: 210000
+
+VERSION: 2

app.py

@@ -0,0 +1,73 @@
+try:
+    import detectron2
+except:
+    import os 
+    os.system('pip install git+https://github.com/facebookresearch/detectron2.git')
+import gradio as gr
+import torch
+import cv2
+import pandas as pd
+from detectron2.utils.visualizer import Visualizer
+from detectron2.data import MetadataCatalog
+from detectron2.config import get_cfg
+from detectron2.data.detection_utils import read_image
+from detectron2.engine import DefaultPredictor
+cfg = get_cfg()
+cfg.merge_from_file("resnet.yaml")
+cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.7
+cfg.MODEL.DEVICE='cpu'
+predictor = DefaultPredictor(cfg)
+
+def inference(image_path):
+    # load image
+    image = read_image(image_path, format="BGR")
+
+    # Make inference on the image
+    outputs = predictor(image)
+
+    metadata = MetadataCatalog.get("custom_dataset_train")
+    MetadataCatalog.get("custom_dataset_train").set(thing_classes=['100km/h', '120km/h', '20km/h', '30km/h', '40km/h', '50km/h', '60km/h', '70km/h', '80km/h'])
+    v = Visualizer(image, metadata, scale=1)
+    out = v.draw_instance_predictions(outputs['instances'])
+
+    # Detection summary table
+    cls_idxs = outputs['instances'].pred_classes.numpy()
+    thing_classes=['100km/h', '120km/h', '20km/h', '30km/h', '40km/h', '50km/h', '60km/h', '70km/h', '80km/h']
+    # get labels from class indices
+    labels = [thing_classes[i] for i in cls_idxs]
+    scores = outputs['instances'].scores.numpy()
+    df = pd.DataFrame({'Detected speed limit': labels, 'Confidence score': scores})
+    # Return the visualization as an RGB image
+    return out.get_image()[:, :, ::-1], df
+
+
+
+examples = ["examples/1.jpg", "examples/2.jpg", "examples/3.jpg"]
+with gr.Blocks(theme='gradio/monochrome') as demo:
+    gr.Markdown("# Speed Limit Detection demo")
+    gr.Markdown("**Author**: *Lu CHEN*")
+    gr.Markdown(
+        """This interactive demo is based on the Faster R-CNN model for object detection. The model is 
+        trained using the [Detectron2](https://github.com/facebookresearch/detectron2) library with a custom
+          dataset that I created by combining images from [Tsinghua-Tencent100K](https://cg.cs.tsinghua.edu.cn/traffic-sign/) and [GTSDB](https://benchmark.ini.rub.de/), both of which provide real-world traffic signs captured within the autonomous driving domain.
+
+To use the demo, simply upload an image and click on *"Infer"* to view the following results: 
+- **Detection**: outputs of Object Detector
+- **Detection summary**: a summary of the detection outputs
+
+You can also select an image from the cached **Examples** to quickly try out. Without clicking *"Infer"*, the cached outputs will be loaded automatically. 
+In case the output image seems too small, simply right-click on the image, and choose “Open image in new tab” to visualize it in full size.
+    """
+    )
+    with gr.Row():
+        with gr.Column():
+            image = gr.Image(type="filepath")
+            button = gr.Button("Infer")
+            
+        with gr.Column():
+            detection = gr.Image(label="Output")
+            detection_summary = gr.DataFrame(label="Detection summary")
+    examples_block = gr.Examples(inputs=image, examples=examples, fn=inference, outputs=[detection, detection_summary], cache_examples=True)
+    button.click(fn=inference, inputs=image, outputs=[detection, detection_summary])
+
+demo.launch()

model_final.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a775a23a4481e1b963650cf26af9db8f8ad3f00724e03485594c8acef239581
+size 330352803

requirements.txt

@@ -0,0 +1,6 @@
+opencv-python-headless
+pandas
+numpy
+matplotlib
+torch
+torchvision

resnet.yaml

@@ -0,0 +1,28 @@
+_BASE_: "Base-RCNN-FPN.yaml"
+MODEL:
+  META_ARCHITECTURE: "GeneralizedRCNN"
+  WEIGHTS: "model_final.pth"
+  #WEIGHTS: "./data/SPEED-Detection/faster-rcnn/vanilla/random_seed_0/model_final.pth"
+
+  MASK_ON: False
+  RESNETS:
+    DEPTH: 50
+  ROI_HEADS:
+    NAME: "StandardROIHeads"
+    NUM_CLASSES: 9
+# INPUT:
+#   MIN_SIZE_TRAIN: (480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800)
+#   MIN_SIZE_TEST: 800
+DATASETS:
+  TRAIN: ('speed_custom_train',)
+  TEST: ('speed_custom_val',)
+SOLVER:
+  IMS_PER_BATCH: 18
+  BASE_LR: 0.01
+  STEPS: (2800, 3700)
+  MAX_ITER: 4200
+  WARMUP_ITERS: 50
+  CHECKPOINT_PERIOD : 2100
+TEST:
+  EVAL_PERIOD: 1000
+