javascriptpythontodataurlbytesio

A image converted with canvas fails to be read by python PIL _io.BytesIO


Python PIL rejects to read an image you have resized with Javascript canvas

I resize an image on the client-side with Javascript:

var reader = new FileReader();
    reader.onload = function (e) {
        el('image-picked').src = e.target.result;
        el('image-picked').className = '';
        var image = new Image();

        //compress Image
        image.onload=function(){
            el("image-picked").src=image.src;
            var canvas=document.createElement("canvas");
            var context=canvas.getContext("2d");
            new_size = get_sizes(image.width,image.height,max_side_px)
            [canvas.width,canvas.height] = new_size;
            context.drawImage(image,
                0,
                0,
                image.width,
                image.height,
                0,
                0,
                canvas.width,
                canvas.height
            );
            console.log("Converted");

            //el('image-picked').className = 'no-display'
            //el('image-picked').src=""
            el('upload-Preview').className = ''
            el('upload-Preview').src = canvas.toDataURL("image/png", quality);

The result seems ok, less size, seemingly ok: enter image description here There are only minor differences on the files with identify: enter image description here

(base) ➜  Desktop file before.png after.png
before.png: PNG image data, 4048 x 3036, 8-bit/color RGB, non-interlaced
after.png:  PNG image data, 500 x 375, 8-bit/color RGBA, non-interlaced

Then I send the file via POST:

    var xhr = new XMLHttpRequest();
    var loc = window.location
    xhr.open('POST', `${loc.protocol}//${loc.hostname}:${loc.port}/analyze`, true);
    xhr.onerror = function() {alert (xhr.responseText);}
    xhr.onload = function(e) {
        if (this.readyState === 4) {
            var response = JSON.parse(e.target.responseText);
            el('result-label').innerHTML = `Result = ${response['result']}`;
        }
    }

    var fileData = new FormData();
    var file = new File([el('upload-Preview').src],
      "image.png", { type: "image/png",
                     lastModified : new Date()});
    fileData.append('file', uploadFiles[0]);
    xhr.send(fileData);

And then I read on the server with python open_image(BytesIO(img_bytes)) :

@app.route('/analyze', methods=['POST'])
async def analyze(request):
    data = await request.form()
    img_bytes = await (data['file'].read())
    img = open_image(BytesIO(img_bytes))

The above works without problems with any normal image, but it fails with any image that is the result of the resize with js, and the error is

File "/Users/brunosan/anaconda3/envs/fastai/lib/python3.7/site-packages/PIL/Image.py", line 2705, in open
    % (filename if filename else fp))
OSError: cannot identify image file <_io.BytesIO object at 0x124ce6360>```

I've tried canvas.toDataURL("image/jpeg", quality) on the JS side, and reading directly with PIL (not fastai, which calls PIL). It's the same error :frowning_face:


Solution

  • Found the answer here.

    I was injecting the image as a a DataURL, when the POST expected a binary. I could see the difference using the "Network" tab: enter image description here

    To convert a DataURL into the binary we need to make a Blob, and then put it into a File:

    function dataURItoBlob(dataURI) {
        // convert base64/URLEncoded data component to raw binary data held in a string
        var byteString;
        if (dataURI.split(',')[0].indexOf('base64') >= 0)
            byteString = atob(dataURI.split(',')[1]);
        else
            byteString = unescape(dataURI.split(',')[1]);
    
        // separate out the mime component
        var mimeString = dataURI.split(',')[0].split(':')[1].split(';')[0];
    
        // write the bytes of the string to a typed array
        var ia = new Uint8Array(byteString.length);
        for (var i = 0; i < byteString.length; i++) {
            ia[i] = byteString.charCodeAt(i);
        }
    
        return new Blob([ia], {type:mimeString});
    }
    
    blob=dataURItoBlob(el('upload-Preview').src)
    var file = new File([blob],
          "image.png", { type: "image/png",
                         lastModified : new Date()});
    var fileData = new FormData();
    fileData.append('file', file);