Streamline Model Inference Pipeline and Checkpoint Management

README.md

@@ -338,6 +338,11 @@ python scripts/gradio_demo.py \
 > [!TIP]
 > Live inference is provided as an independent script so that you can load this script into any deployment devices without having to install all the (large) dependencies in this repo. The script only requires `onnxruntime`, `cv2` and `numpy` to run.
 
+There are two live inference scripts:
+
+1. `live_inference.py` - Run live inference on a video, image or webcam using ONNXRuntime. This script requires a pre-exported ONNX model.
+2. `live_inference_pretrained.py` - Automatically download the pretrained model, convert to ONNX, and run inference in one go.
+
 Run live inference on a video, image or webcam using ONNXRuntime. This runs on CPU by default.
 If you would like to use the CUDA backend, install the `onnxruntime-gpu` package and uninstall the `onnxruntime` package.
 
@@ -378,6 +383,18 @@ Webcam video width at 320x240 pixels (240p):
 https://github.com/user-attachments/assets/f4afff9c-3e6d-4965-ab86-0d4de7ce1a44
 
 
+To run live inference with pretrained model on webcam, use the following command
+```bash
+python scripts/live_inference_pretrained.py --webcam
+```
+This downloads the pretrained model, converts it to ONNX, and runs inference on the webcam. All other arguments are optional but they are similar to the `live_inference.py` script.
+
+The output is as follows
+
+
+https://github.com/user-attachments/assets/5aca6044-db81-4988-937e-2447a7cbe06e
+
+
 
 
 For video inference, specify the path to the video file as the input. Output video will be saved as `onnx_result.mp4` in the current directory.
@@ -393,7 +410,12 @@ python scripts/live_inference.py
 ```
 https://github.com/user-attachments/assets/6bc1dc6a-a223-4220-954d-2dab5c75b4a8
 
-The following is an inference using the pre-trained model `deim_hgnetv2_x` trained on COCO. See how I exported the pre-trained model to onnx in this notebook [here](nbs/export.ipynb).
+
+The following is an inference using the pre-trained model `deim_hgnetv2_x` trained on COCO.
+
+```bash
+python scripts/live_inference_pretrained.py --model deim_hgnetv2_x --video video.mp4
+```
 
 https://github.com/user-attachments/assets/77070ea4-8407-4648-ade3-01cacd77b51b
 
@@ -473,6 +495,11 @@ pixi run -e cuda live-inference --onnx model.onnx --webcam --provider cuda --cla
 pixi run -e cpu live-inference --onnx model.onnx --input video.mp4 --class-names classes.txt --inference-size 320
 ```
 
+Live inference with pretrained model on webcam
+```bash
+pixi run -e cuda live-inference-pretrained --webcam
+```
+
 Launch Gradio app
 ```bash
 pixi run gradio-demo --model "best_prep.onnx" --classes "classes.txt" --examples "Rock Paper Scissors SXSW.v14i.coco/test"

pyproject.toml

@@ -11,7 +11,6 @@ dependencies = [
     "calflops>=0.3.2,<0.4",
     "transformers>=4.49.0,<5",
     "opencv-python>=4.11.0.86,<5",
-    "gdown>=5.2.0,<6",
     "loguru>=0.7.3,<0.8",
     "matplotlib",
     "onnx>=1.17.0,<2",
@@ -66,6 +65,7 @@ default = { features = ["cpu"] }
 [tool.pixi.tasks]
 quickstart = "python scripts/quickstart.py"
 live-inference = "python scripts/live_inference.py"
+live-inference-pretrained = "python scripts/live_inference_pretrained.py"
 train-model = "python scripts/train.py"
 gradio-demo = "python scripts/gradio_demo.py"
 export = "python scripts/export.py"