add video processing sample

Author: ks6088ts

apps/99_streamlit_examples/README.md

@@ -77,6 +77,10 @@ Access to http://localhost:8501 and select the sample you want to run from the s
 
 ![Pose estimation](../../docs/images/99_streamlit_examples.poseestimation.png)
 
+#### 12. Video processing
+
+![Video processing](../../docs/images/99_streamlit_examples.videoprocessing.png)
+
 ## References
 
 - [🎈 Streamlit + LLM Examples App](https://github.com/streamlit/llm-examples)

apps/99_streamlit_examples/pages/12_Video_processing.py

@@ -0,0 +1,81 @@
+import cv2
+import streamlit as st
+from dotenv import load_dotenv
+from ultralytics import YOLO
+
+load_dotenv()
+
+
+class Processor:
+    def __init__(self, model_name):
+        self.model = YOLO(model_name)
+
+    def process(self, frame):
+        results = self.model(
+            frame,
+            conf=0.5,
+            classes=[0],
+        )
+        output_img = results[0].plot(
+            labels=True,
+            conf=True,
+        )
+        return cv2.cvtColor(
+            src=output_img,
+            code=cv2.COLOR_BGR2RGB,
+        )
+
+
+with st.sidebar:
+    model_name = st.selectbox(
+        label="Select a model",
+        options=[
+            "yolov8n.pt",
+            "yolov9c.pt",
+            "yolov10n.pt",
+            # https://docs.ultralytics.com/models/yolov8/#supported-tasks-and-modes
+        ],
+        key="model_name",
+        index=0,
+    )
+    device = st.text_input(
+        label="input your video/camera device",
+        value="0",
+    )
+    if device.isnumeric():
+        # e.g. "0" -> 0
+        device = int(device)
+
+st.title("Video processing")
+
+start_button = st.button("Start")
+stop = st.button("Stop")
+
+image_loc = st.empty()
+processor = Processor(
+    model_name=model_name,
+)
+
+if start_button:
+    capture = cv2.VideoCapture(device)
+
+    while capture.isOpened:
+        ret, frame = capture.read()
+
+        if not ret:
+            st.error("Failed to capture image")
+            continue
+
+        processed_frame = processor.process(
+            frame=frame,
+        )
+
+        image_loc.image(
+            image=processed_frame,
+            use_column_width=True,
+        )
+
+        if stop:
+            break
+
+    capture.release()

docs/images/99_streamlit_examples.videoprocessing.png

@@ -25,6 +25,7 @@ mediapipe = "^0.10.14"
 azure-ai-documentintelligence = "^1.0.0b3"
 azure-storage-blob = "^12.22.0"
 requests = "^2.32.3"
+opencv-python-headless = "^4.10.0.84"
 
 [tool.poetry.group.dev.dependencies]
 pre-commit = "^3.8.0"

poetry.lock

@@ -9,11 +9,15 @@ langchain-openai==0.1.20
 langchain-community==0.2.11
 azure-search-documents==11.5.1
 azure-identity==1.17.1
+azure-ai-documentintelligence==1.0.0b3
+azure-storage-blob==12.22.0
+requests==2.32.3
 
 # To run 99_streamlit_examples/pages/10_Object_Detection.py
 # ultralytics==8.2.77
 
 # To run 99_streamlit_examples/pages/11_Pose_Estimation.py
 # mediapipe==0.10.14
 
-azure-ai-documentintelligence==1.0.0b3
+# To run 99_streamlit_examples/pages/12_Video_processing.py
+# opencv-python-headless==4.10.0.84

pyproject.toml

@@ -23,6 +23,7 @@ def test_smoke():
         # fixme: disabled due to flaky test
         # "apps/99_streamlit_examples/pages/10_Object_detection.py",
         # "apps/99_streamlit_examples/pages/11_Pose_estimation.py",
+        "apps/99_streamlit_examples/pages/12_Video_processing.py",
     ]
     for path in paths:
         at = AppTest(