155 lines
6.6 KiB
React
155 lines
6.6 KiB
React
import { useState } from 'react'
|
|
import { useRef } from "react";
|
|
import './Bird_CNN.css';
|
|
|
|
function Bird_CNN() {
|
|
const apiUrl = process.env.NODE_ENV === "development"
|
|
? "https://marvinkrausser.com/cnn-api"
|
|
: "https://marvinkrausser.com/cnn-api";
|
|
|
|
const [file, setFile] = useState(null);
|
|
const [birdClass, setBirdClass] = useState(null);
|
|
const [confidence, setConfidence] = useState(null);
|
|
const [error, setError] = useState(false);
|
|
const [preview, setPreview] = useState("/cherry_bird.jpeg");
|
|
const [loading, setLoading] = useState(false);
|
|
const fileInputRef = useRef(null);
|
|
const uploadButton = useRef(null);
|
|
const scrollRefClassifiction = useRef(null);
|
|
const scrollRefUploadButton = useRef(null);
|
|
|
|
const isPartiallyInViewport = (el) => {
|
|
if (!el) return false;
|
|
|
|
const rect = el.getBoundingClientRect();
|
|
|
|
return (
|
|
rect.top < window.innerHeight &&
|
|
rect.bottom > 0
|
|
);
|
|
};
|
|
|
|
const handleImageDivClick = () => {
|
|
fileInputRef.current.click();
|
|
}
|
|
|
|
const handleImage = (e) => {
|
|
if (!e.target.files[0]) { return; }
|
|
setFile(e.target.files[0]);
|
|
setPreview(URL.createObjectURL(e.target.files[0]));
|
|
uploadButton.current.classList.remove("inactive");
|
|
|
|
scrollRefUploadButton.current.scrollIntoView({ behavior: "smooth" });
|
|
|
|
setConfidence(null);
|
|
setBirdClass(null);
|
|
|
|
};
|
|
|
|
const sendImage = async (e) => {
|
|
if (!file) return;
|
|
|
|
scrollRefClassifiction.current.scrollIntoView({ behavior: "smooth" });
|
|
|
|
const formData = new FormData();
|
|
formData.append("file", file);
|
|
|
|
setLoading(true);
|
|
|
|
try {
|
|
const response = await fetch(`${apiUrl}/predict`, {
|
|
method: "POST",
|
|
body: formData,
|
|
});
|
|
} catch (e) {
|
|
setError(true);
|
|
return;
|
|
}
|
|
finally {
|
|
setLoading(false);
|
|
}
|
|
|
|
if (!response.ok) {
|
|
setError(true);
|
|
return;
|
|
}
|
|
else {
|
|
setError(false);
|
|
}
|
|
|
|
const result = await response.json();
|
|
setBirdClass(result["class"]);
|
|
const confidence = result["confidence"];
|
|
setConfidence(`${Math.round(confidence * 100)}%`);
|
|
};
|
|
|
|
return (
|
|
<>
|
|
<div className='site-box' style={{ marginTop: "150px" }}>
|
|
<h1 id='site-headline'>Bird Species Expert</h1>
|
|
<div className='content-box' style={{ display: "flex", flexWrap: "wrap", justifyContent: "center" }}>
|
|
<div className='explanation-box left'>
|
|
<h2 style={{ color: "rgb(47, 168, 208)" }}>Explanation</h2>
|
|
<span>Select an image and upload it to our bird expert. You will receive a classification and how certain the expert is with her opinion. Be aware that the expert may not be always right.</span>
|
|
</div>
|
|
|
|
<div className='request-box' style={{ height: "800px", width: "700px", display: "flex", flexDirection: "column", alignItems: "center", flex: "0 0 auto", margin: "100px 50px 0px 50px" }}>
|
|
<div ref={scrollRefUploadButton} className='input-box' style={{ scrollMarginTop: "100px", marginBottom: "80px", width: "500px", height: "40px", display: "flex", justifyContent: "space-evenly", alignItems: "center" }}>
|
|
<div>
|
|
<input disabled={loading} ref={fileInputRef} type="file" id='fileUpload' accept="image/jpeg" onChange={handleImage} style={{ display: "none" }} />
|
|
<label htmlFor="fileUpload" className="custom-button">
|
|
Select Image
|
|
</label>
|
|
</div>
|
|
<div>
|
|
<button id='button-send' onClick={sendImage} style={{ display: "none" }} disabled={loading} />
|
|
<label htmlFor="button-send" className="custom-button inactive" ref={uploadButton}>
|
|
Ask Expert
|
|
</label>
|
|
</div>
|
|
|
|
</div>
|
|
|
|
<div className='image-box'>
|
|
<img
|
|
onClick={handleImageDivClick}
|
|
disabled={loading}
|
|
src={preview}
|
|
alt="preview"
|
|
style={{ height: "100%", userSelect: "none", maxWidth: "700px" }}
|
|
/>
|
|
</div>
|
|
|
|
<div ref={scrollRefClassifiction} style={{ scrollMarginTop: "100px", paddingTop: "10px", marginTop: "30px", width: "85px", height: "35px", display: "flex", justifyContent: "center", minHeight: "25px", minWidth: "85px" }}>
|
|
{loading && <div className='loader'></div>}
|
|
</div>
|
|
|
|
|
|
<div className='response-block' style={{ marginTop: "40px", width: "80%", margin: "40px", display: "flex", flexDirection: "column", alignItems: "flex-start", gap: "10px" }}>
|
|
<div className='content-block class'>
|
|
<h3>Bird Species: </h3>
|
|
<p id='bird-class-text'>{birdClass}</p>
|
|
</div>
|
|
<div className='content-block confidence'>
|
|
<h4>Model Confidence: </h4>
|
|
<p id='bird-confidence-text'>{confidence}</p>
|
|
</div>
|
|
{error && <h4>An Error has uccured. Please try again later.</h4>}
|
|
</div>
|
|
</div>
|
|
|
|
<div className='explanation-box right'>
|
|
<h3>Model Architecture</h3>
|
|
<div style={{ display: "inline" }}>
|
|
<span>The model used for classification is a convolutional neural network (CNN) based on depthwise separable convolutions, as introduced in the </span>
|
|
<a href='https://arxiv.org/pdf/1610.02357' target='_blank' rel='noopener noreferrer'>Xception: Deep Learning with Depthwise Separable Convolutions</a>
|
|
<span> paper by François Chollet. The model uses who knows how many layers, a dropout and multiple batchnormalsation.</span>
|
|
</div>
|
|
</div>
|
|
</div>
|
|
</div >
|
|
</>
|
|
);
|
|
}
|
|
|
|
export default Bird_CNN; |