Using annotation form data to fill HTML value

, ,
1

Hello,

I'm trying to build an pdf annotation tool page by page and I want to add some html to show the progression. I save on the input stream the number of the page and some information. I want to access the data inside the stream but it does't work.

Here is the code:

@prodigy.recipe("classify-images.manual")
def classify_images_manual(dataset, source):
    def get_stream():
        files = os.listdir(source)

        for f in files:
            # Load each PDF and save images
            pages = convert_from_path(f"{source}/{f}")

            for i, p in enumerate(pages):
                file_name = f.split(".")[0]
                image_file_name = f"{os.getcwd()}/images/{file_name}_{i}.jpg"
                p.save(image_file_name, "JPEG")

                im = {"image": image_file_name}
                im["options"] = OPTIONS
                im["file_name"] = file_name
                im["page"] = i + 1
                im["total"] = len(pages)

                yield im

    return {
        "dataset": dataset,
        "stream": fetch_images(get_stream()),
        "view_id": "blocks",
        "config": {
            "blocks": [
                {
                    "view_id": "html",
                    "html": "<h1>File {{file_name}} ({{page}}/{{total}})</h1>",
                },
                {"view_id": "choice", "choice_style": "single",},  # or "multiple"
            ]
        },
    }

I don't know if it has any relation with the fetch_images function.

2

Hi! The fetch_images processor only accesses the "image" value of a task and converts that to a base64 string, so this shouldn't make a difference.

I think the problem here is that the variables need to be defined in a html_template, which is basically an alternative to hard-coding the html string in your task. So if you change html to html_template in your blocks, it should work.

Alternatively, you could also just generate the HTML in your stream, e,g,:

im["html"] = f"<h1>File {file_name}</h1>"  # etc.