I’m trying to create a project in Windows that integrates a pre-trained Faster R-CNN model with ResNet-50 and Feature Pyramid Network (FPN) for object detection, using Rust and Python. The program processes images in a specified folder, detects objects in each image, and displays the results in a simple Python GUI application.

My problem is that I can’t get OpenCV to work. What do is:

• I’m creating new project with cargo new object_detection
• Then virtual environment with python -m venv venv
• And then use develop maturin to deploy the package from Rust so that it can be referenced from Python

lib.rs:

use pyo3::prelude::*;
use opencv::prelude::*;
use opencv::imgcodecs;
use tch::vision::detector::{ObjectDetector, VisionTask};
use std::fs;
use rayon::prelude::*;

#[pyfunction]
fn process_images(folder_path: &str) -> PyResult<String> {
    // Load the pre-trained model
    let detector = ObjectDetector::new("./model/fasterrcnn_resnet50_fpn.pth")
        .map_err(|e| pyo3::exceptions::PyRuntimeError::new_err(e.to_string()))?;
        
    let paths: Vec<_> = fs::read_dir(folder_path)
        .map_err(|e| pyo3::exceptions::PyRuntimeError::new_err(e.to_string()))?
        .filter_map(|entry| entry.ok())
        .collect();

    let results: Vec<String> = paths.par_iter().map(|path| {
        let path = path.path();
        if path.is_file() {
            match imgcodecs::imread(path.to_str().unwrap(), imgcodecs::IMREAD_COLOR) {
                Ok(img) => {
                    // Convert OpenCV Mat to Tensor
                    let tensor = tch::vision::image::load(path.to_str().unwrap()).unwrap();
                    match detector.detect(&tensor) {
                        Ok(detections) => format!("Results for {}: {:?}n", path.display(), detections),
                        Err(e) => format!("Error detecting objects in {}: {}n", path.display(), e),
                    }
                },
                Err(e) => format!("Error reading image {}: {}n", path.display(), e),
            }
        } else {
            format!("{} is not a filen", path.display())
        }
    }).collect();

    Ok(results.concat())
}

#[pymodule]
fn object_detection(_py: Python, m: &PyModule) -> PyResult<()> {
    m.add_function(wrap_pyfunction!(process_images, m)?)?;
    Ok(())
}

GUI.py:

import tkinter as tk
from tkinter import filedialog
import object_detection

def upload_folder():
    folder_path = filedialog.askdirectory()
    if folder_path:
        process_images(folder_path)

def process_images(folder_path):
    results = object_detection.process_images(folder_path)
    display_results(results)

def display_results(results):
    result_text.delete(1.0, tk.END)
    result_text.insert(tk.END, results)

root = tk.Tk()
root.title("Object Detection Application")

upload_button = tk.Button(root, text="Upload Folder", command=upload_folder)
upload_button.pack()

result_text = tk.Text(root)
result_text.pack()

root.mainloop()

And when I try to create package with maturin develop I get an error: note: LINK : fatal error LNK1181: cannot open input file 'libclang.lib'.

Before that I installed sufficient packages with choco for Windows:

Set-ExecutionPolicy Bypass -Scope Process -Force; [System.Net.ServicePointManager]::SecurityProtocol = [System.Net.ServicePointManager]::SecurityProtocol -bor 3072; iex ((New-Object System.Net.WebClient).DownloadString('https://community.chocolatey.org/install.ps1'))

And then:

choco install llvm opencv

I explicitly created environment variables:

LIBCLANG_PATH = C:Program FilesLLVMlib
OPENCV_INCLUDE_PATHS = C:toolsopencvbuildinclude
OPENCV_LINK_LIBS = opencv_world490
OPENCV_LINK_PATHS = C:toolsopencvbuildx64vc16lib

My Cargo.toml looks like this:

[package]
name = "object_detection"
version = "0.1.0"
edition = "2021"

[dependencies]
pyo3 = { version = "0.15", features = ["extension-module"] }
opencv = "0.92.0"
tch = "0.1"
rayon = "1.5"

[lib]
crate-type = ["cdylib"]

It doesn’t work with the same error after building this package with cardo build.

What might be the issue?