yl-ocr-layout/server.py

from fastapi import FastAPI, Request, Form, File, UploadFile
from fastapi.templating import Jinja2Templates
from pydantic import BaseModel
from typing import List, Optional
from sx_utils import web_try
import cv2
import numpy as np

import torch
import base64
import random

# YOLO_DIR = '/workspace/yolov5'
YOLO_DIR = '/Users/sxkj/opt/python-workspace/yili-ocr/yl-ocr-layout/yolov5'
# WEIGHTS = '/data/yolov5/runs/train/yolov5x_layout_reuslt37/weights/best.pt'
# WEIGHTS = '/workspace/best.pt'
WEIGHTS = '/Users/sxkj/opt/python-workspace/yili-ocr/yl-ocr-layout/best.pt'

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")


app = FastAPI()
templates = Jinja2Templates(directory = 'templates')

model_selection_options = ['ocr-layout']
model_dict = {model_name: None for model_name in model_selection_options} #set up model cache

colors = [tuple([random.randint(0, 255) for _ in range(3)]) for _ in range(100)] #for bbox plotting

##############################################
#-------------GET Request Routes--------------
##############################################
@app.get("/")
def home(request: Request):
    ''' Returns html jinja2 template render for home page form
    '''

    return templates.TemplateResponse('home.html', {
            "request": request,
            "model_selection_options": model_selection_options,
        })

@app.get("/drag_and_drop_detect")
def drag_and_drop_detect(request: Request):
    ''' drag_and_drop_detect detect page. Uses a drag and drop
    file interface to upload files to the server, then renders
    the image + bboxes + labels on HTML canvas.
    '''

    return templates.TemplateResponse('drag_and_drop_detect.html',
            {"request": request,
            "model_selection_options": model_selection_options,
        })


##############################################
#------------POST Request Routes--------------
##############################################
@app.post("/")
def detect_via_web_form(request: Request,
                        file_list: List[UploadFile] = File(...),
                        model_name: str = Form(...),
                        img_size: int = Form(1824)):

    '''
    Requires an image file upload, model name (ex. yolov5s). Optional image size parameter (Default 1824).
    Intended for human (non-api) users.
    Returns: HTML template render showing bbox data and base64 encoded image
    '''

    #assume input validated properly if we got here
    if model_dict[model_name] is None:
        model_dict[model_name] = model = torch.hub.load(YOLO_DIR, 'custom', path=WEIGHTS, source='local').to(device)

    img_batch = [cv2.imdecode(np.fromstring(file.file.read(), np.uint8), cv2.IMREAD_COLOR)
                    for file in file_list]

    #create a copy that corrects for cv2.imdecode generating BGR images instead of RGB
    #using cvtColor instead of [...,::-1] to keep array contiguous in RAM
    img_batch_rgb = [cv2.cvtColor(img, cv2.COLOR_BGR2RGB) for img in img_batch]

    results = model_dict[model_name](img_batch_rgb, size = img_size)

    json_results = results_to_json(results,model_dict[model_name])

    img_str_list = []
    #plot bboxes on the image
    for img, bbox_list in zip(img_batch, json_results):
        for bbox in bbox_list:
            label = f'{bbox["class_name"]} {bbox["confidence"]:.2f}'
            plot_one_box(bbox['bbox'], img, label=label,
                    color=colors[int(bbox['class'])], line_thickness=3)

        img_str_list.append(base64EncodeImage(img))

    #escape the apostrophes in the json string representation
    encoded_json_results = str(json_results).replace("'",r"\'").replace('"',r'\"')

    return templates.TemplateResponse('show_results.html', {
            'request': request,
            'bbox_image_data_zipped': zip(img_str_list,json_results), #unzipped in jinja2 template
            'bbox_data_str': encoded_json_results,
        })


@app.post("/detect")
@web_try()
def detect_via_api(request: Request,
                file_list: List[UploadFile] = File(...),
                model_name: str = Form(...),
                img_size: Optional[int] = Form(1824),
                download_image: Optional[bool] = Form(False)):

    '''
    Requires an image file upload, model name (ex. yolov5s).
    Optional image size parameter (Default 1824)
    Optional download_image parameter that includes base64 encoded image(s) with bbox's drawn in the json response

    Returns: JSON results of running YOLOv5 on the uploaded image. If download_image parameter is True, images with
            bboxes drawn are base64 encoded and returned inside the json response.

    Intended for API usage.
    '''

    if model_dict[model_name] is None:
          model_dict[model_name] = model = torch.hub.load(YOLO_DIR, 'custom', path=WEIGHTS, source='local').to(device)

    img_batch = [cv2.imdecode(np.fromstring(file.file.read(), np.uint8), cv2.IMREAD_COLOR)
                for file in file_list]

    #create a copy that corrects for cv2.imdecode generating BGR images instead of RGB,
    #using cvtColor instead of [...,::-1] to keep array contiguous in RAM
    img_batch_rgb = [cv2.cvtColor(img, cv2.COLOR_BGR2RGB) for img in img_batch]

    results = model_dict[model_name](img_batch_rgb, size = img_size)
    json_results = results_to_json(results,model_dict[model_name])

    if download_image:
        for idx, (img, bbox_list) in enumerate(zip(img_batch, json_results)):
            for bbox in bbox_list:
                label = f'{bbox["class_name"]} {bbox["confidence"]:.2f}'
                plot_one_box(bbox['bbox'], img, label=label,
                        color=colors[int(bbox['class'])], line_thickness=3)

            payload = {'image_base64':base64EncodeImage(img)}
            json_results[idx].append(payload)

    encoded_json_results = str(json_results).replace("'",r'"')
    return encoded_json_results

##############################################
#--------------Helper Functions---------------
##############################################

def results_to_json(results, model):
    ''' Converts yolo model output to json (list of list of dicts)'''
    return [
                [
                    {
                    "class": int(pred[5]),
                    "class_name": model.model.names[int(pred[5])],
                    "bbox": [int(x) for x in pred[:4].tolist()], #convert bbox results to int from float
                    "confidence": float(pred[4]),
                    }
                for pred in result
                ]
            for result in results.xyxy
            ]


def plot_one_box(x, im, color=(128, 128, 128), label=None, line_thickness=3):
    # Directly copied from: https://github.com/ultralytics/yolov5/blob/cd540d8625bba8a05329ede3522046ee53eb349d/utils/plots.py
    # Plots one bounding box on image 'im' using OpenCV
    assert im.data.contiguous, 'Image not contiguous. Apply np.ascontiguousarray(im) to plot_on_box() input image.'
    tl = line_thickness or round(0.002 * (im.shape[0] + im.shape[1]) / 2) + 1  # line/font thickness
    c1, c2 = (int(x[0]), int(x[1])), (int(x[2]), int(x[3]))
    cv2.rectangle(im, c1, c2, color, thickness=tl, lineType=cv2.LINE_AA)
    if label:
        tf = max(tl - 1, 1)  # font thickness
        t_size = cv2.getTextSize(label, 0, fontScale=tl / 3, thickness=tf)[0]
        c2 = c1[0] + t_size[0], c1[1] - t_size[1] - 3
        cv2.rectangle(im, c1, c2, color, -1, cv2.LINE_AA)  # filled
        cv2.putText(im, label, (c1[0], c1[1] - 2), 0, tl / 3, [225, 255, 255], thickness=tf, lineType=cv2.LINE_AA)


def base64EncodeImage(img):
    ''' Takes an input image and returns a base64 encoded string representation of that image (jpg format)'''
    _, im_arr = cv2.imencode('.jpg', img)
    im_b64 = base64.b64encode(im_arr.tobytes()).decode('utf-8')

    return im_b64

if __name__ == '__main__':
    import uvicorn
    import argparse
    parser = argparse.ArgumentParser()
    parser.add_argument('--host', default = 'localhost')
    parser.add_argument('--port', default = 8080)
    parser.add_argument('--precache-models', action='store_true',
            help='Pre-cache all models in memory upon initialization, otherwise dynamically caches models')
    opt = parser.parse_args()

    # if opt.precache_models:
    #     model_dict = {model_name: torch.hub.load('ultralytics/yolov5', model_name, pretrained=True)
    #                     for model_name in model_selection_options}

    app_str = 'server:app' #make the app string equal to whatever the name of this file is
    uvicorn.run(app_str, host= opt.host, port=int(opt.port), reload=True)