init update

.gitattributes

@@ -1,35 +1,39 @@
 *.7z filter=lfs diff=lfs merge=lfs -text
 *.arrow filter=lfs diff=lfs merge=lfs -text
 *.bin filter=lfs diff=lfs merge=lfs -text
+*.bin.* filter=lfs diff=lfs merge=lfs -text
 *.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
 *.ftz filter=lfs diff=lfs merge=lfs -text
 *.gz filter=lfs diff=lfs merge=lfs -text
 *.h5 filter=lfs diff=lfs merge=lfs -text
 *.joblib filter=lfs diff=lfs merge=lfs -text
 *.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
 *.model filter=lfs diff=lfs merge=lfs -text
 *.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
 *.onnx filter=lfs diff=lfs merge=lfs -text
 *.ot filter=lfs diff=lfs merge=lfs -text
 *.parquet filter=lfs diff=lfs merge=lfs -text
 *.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
 *.pt filter=lfs diff=lfs merge=lfs -text
 *.pth filter=lfs diff=lfs merge=lfs -text
 *.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
 saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
 *.tflite filter=lfs diff=lfs merge=lfs -text
 *.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
 *.xz filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.zstandard filter=lfs diff=lfs merge=lfs -text
+*.tfevents* filter=lfs diff=lfs merge=lfs -text
+*.db* filter=lfs diff=lfs merge=lfs -text
+*.ark* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*data* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.meta filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.index filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.gguf* filter=lfs diff=lfs merge=lfs -text
+*.ggml filter=lfs diff=lfs merge=lfs -text
+*.llamafile* filter=lfs diff=lfs merge=lfs -text
+test_imgs/* filter=lfs diff=lfs merge=lfs -text
+demo/gallery/* filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,3 @@
+demo/__pycache__
+workdir/
+.DS_Store

app.py

@@ -0,0 +1,145 @@
+import cv2, os
+import gradio as gr
+import numpy as np
+from demo.generation import call_generation
+from demo.instructions import INSTRUCTIONS_VECTORIZE_SIMPLIFY
+
+VERSION = 'v0.1'
+
+GALLERY_LIST = [os.path.join('demo/gallery',path) for path in os.listdir('demo/gallery')]
+
+
+def resize_image(image, size):
+    # find the minimal size of the image, resize it to size
+    # H, W, C = image.shape
+    return cv2.resize(image, (size[0], size[1]), interpolation=cv2.INTER_LINEAR)
+
+def HWC3(x):
+    assert x.dtype == np.uint8
+    if x.ndim == 2:
+        x = x[:, :, None]
+    assert x.ndim == 3
+    H, W, C = x.shape
+    assert C == 1 or C == 3 or C == 4
+    if C == 3:
+        return x
+    if C == 1:
+        return np.concatenate([x, x, x], axis=2)
+    if C == 4:
+        color = x[:, :, 0:3].astype(np.float32)
+        alpha = x[:, :, 3:4].astype(np.float32) / 255.0
+        y = color * alpha + 255.0 * (1.0 - alpha)
+        y = y.clip(0, 255).astype(np.uint8)
+        return y
+
+    
+
+
+def process_vector(input_image, upsample_method, svg_simplify, svg_optimize, trace_mode, subsample_ratio, speckle_removal,sorting_method, sorting_order, use_gpu):
+    print("Processing vector:",upsample_method, svg_simplify, svg_optimize, trace_mode)
+    if input_image is not None:
+        ## save input_image to a temp file
+        
+        ## process the image
+        file_list = call_generation(input_image, 
+                        preprocess=upsample_method, 
+                        simplify=svg_simplify, 
+                        optimize=svg_optimize, 
+                        mode=trace_mode, 
+                        subsample_ratio=subsample_ratio, 
+                        speckle_removal=speckle_removal, 
+                        sorting_method=sorting_method,
+                        sorting_order=sorting_order,
+                        use_gpu=use_gpu)
+        
+        return file_list
+
+
+
+
+
+block = gr.Blocks(
+        title = "VectorizeAnything",
+        theme=gr.themes.Soft(
+            radius_size=gr.themes.sizes.radius_none,
+            text_size=gr.themes.sizes.text_md
+        ),
+        css="css/style.css",
+        ).queue()
+
+with block:
+    state = gr.State(value={
+        'gallery_selected_img_path': None,  # 当前选中的图片路径
+        'gallery_selected_img_path_idx': 0,  # 当前选中的图片路径索引
+    })
+    with gr.Row():
+        gr.HTML(f"""
+                    </br>
+                    <div>
+                        <h1 style="font-size:3rem; "><center>Vectorize Anything: {VERSION} </center></h1>
+                    </div>
+                    </br>
+            """)
+    # tab_0 = gr.Tab(label="Gallery (画廊)")
+    # with tab_0:
+    #     with gr.Row():
+    #         gr.Gallery(label='图像生成结果', value=GALLERY_LIST,show_label=False, elem_id="Gallery", columns=5, height=1000)
+
+
+    tab_3 = gr.Tab(label="IMG to SVG")
+    with tab_3:   
+        with gr.Accordion('🕹Usage', open=True,):
+            with gr.Tabs():
+                gr.HTML(INSTRUCTIONS_VECTORIZE_SIMPLIFY)
+        with gr.Row():
+            with gr.Column():
+                input_image = gr.Image(type="numpy")
+                run_vectorize = gr.Button(value="Vectorize",elem_id="btnVEC")
+
+                with gr.Accordion("Vector options", open=True):
+                    upsample_method = gr.Dropdown(choices=["None", "x4", "x2"], type="value", value="None", label="Upsample Method")
+                    sorting_method = gr.Dropdown(choices=["brightness","area"], type="value", value="brightness", label="Sorting Method")
+                    sorting_order = gr.Dropdown(choices=["ascend","descend"], type="value", value="descend", label="Sorting Order")
+                    trace_mode = gr.Radio(choices=["overlap", "cutout"], type="value", value="overlap", label="Trace Mode")
+                    use_gpu = gr.Checkbox(label='use GPU', value=False, visible=True)
+                    svg_simplify = gr.Checkbox(label='Simplify SVG', value=False, visible=True)
+                    svg_optimize = gr.Checkbox(label='Optimize SVG', value=False, visible=True)
+                    speckle_removal = gr.Checkbox(label='Remove small speckle', value=False)
+                    subsample_ratio = gr.Slider(label="Subsample Ratio", minimum=1, maximum=10000, value=12, step=1, visible=False)
+
+
+            def exp_gen_click():
+                return [gr.Slider(value=512), gr.Slider(value=512)]  # all examples are 512x512, refresh draw_img
+
+            with gr.Column():
+                result_vector_gallery = gr.Gallery(label='Output', show_label=False, elem_id="Gallery_vector")
+            
+
+            
+                with gr.Tab("Image Examples"):
+                    exp_gen_en = gr.Examples(
+                    [
+                        ["test_imgs/demo1.png"],
+                        ["test_imgs/demo2.jpg"],
+                        ["test_imgs/demo3.png"],
+                        ["test_imgs/demo4.png"],
+                        ["test_imgs/demo5.png"],
+                        ["test_imgs/demo6.png"],
+                        ["test_imgs/demo7.png"],
+                        ["test_imgs/demo8.png"],
+                        ["test_imgs/demo9.png"],
+                        ["test_imgs/demo10.png"],
+                        ["test_imgs/demo11.png"],
+                        ["test_imgs/demo12.png"],
+                    ],
+                    [input_image],
+                    examples_per_page=20,
+                    label=''
+                    )
+                    exp_gen_en.dataset.click(exp_gen_click, None)
+
+    
+    vector_ips = [input_image, upsample_method, svg_simplify, svg_optimize, trace_mode, subsample_ratio, speckle_removal,sorting_method, sorting_order, use_gpu]
+    run_vectorize.click(fn=process_vector, inputs=vector_ips, outputs=result_vector_gallery)
+
+block.launch(server_name='0.0.0.0', share=False,debug=True, root_path=f"/{os.getenv('GRADIO_PROXY_PATH')}" if os.getenv('GRADIO_PROXY_PATH') else "")

demo/generation.py

@@ -0,0 +1,131 @@
+import json
+import os
+import time
+
+import gradio as gr
+import requests
+
+from demo.log import logger
+from demo.util import download_svgs, upload_np_2_oss
+
+API_KEY = os.getenv("API_KEY_GENERATION")
+
+def convert_bool_to_str(value):
+    if value:
+        return "True"
+    else:
+        return "False"
+
+def call_generation(input_path, 
+                    preprocess, 
+                    simplify, 
+                    optimize, 
+                    mode, 
+                    subsample_ratio, 
+                    speckle_removal, 
+                    sorting_method,
+                    sorting_order,
+                    use_gpu):
+    
+    ## generate image name based on time stamp
+    time_str = time.strftime("%Y%m%d%H%M%S", time.localtime())
+    img_name = f"upload_{time_str}.png"
+    svg_name = f"result_{time_str}"
+    BATCH_SIZE = 1
+    if simplify:
+        BATCH_SIZE += 1
+    if optimize:
+        BATCH_SIZE += 1
+    img_url = upload_np_2_oss(input_path, name=img_name)
+    simplify = convert_bool_to_str(simplify)
+    optimize = convert_bool_to_str(optimize)
+    speckle_removal = convert_bool_to_str(speckle_removal)
+    use_gpu = convert_bool_to_str(use_gpu)
+    headers = {
+        "Content-Type": "application/json",
+        "Accept": "application/json",
+        "Authorization": f"Bearer {API_KEY}",
+        "X-DashScope-Async": "enable",
+    }
+    data = {
+        "model": "pre-vectorize_anything-2333",
+        "input": {
+            "base_image_url": img_url
+        },
+        "parameters":{
+            "preprocess": preprocess,
+            "mode": mode,
+            "simplify": simplify, 
+            "optimize": optimize,
+            "sorting_method": sorting_method,
+            "sorting_order": sorting_order,
+            "subsample_ratio": subsample_ratio,
+            "speckle_removal": speckle_removal, 
+            "use_GPU": use_gpu
+        }
+    }
+    url_create_task = 'https://poc-dashscope.aliyuncs.com/api/v1/services/vision/image-process/process'
+
+    all_res_ = []
+    REPEAT = 1
+    for _ in range(REPEAT):
+        try:
+            res_ = requests.post(url_create_task, data=json.dumps(data), headers=headers, timeout=60)
+            print(json.dumps(data))
+            all_res_.append(res_)
+        except requests.Timeout:
+            # back off and retry
+            raise gr.Error("网络波动，请求失败，请再次尝试")
+
+    all_image_data = []
+    for res_ in all_res_:
+        respose_code = res_.status_code
+        if 200 == respose_code:
+            res = json.loads(res_.content.decode())
+            request_id = res['request_id']
+            task_id = res['output']['task_id']
+            logger.info(f"task_id: {task_id}: Create Poster Imitation request success. Params: {data}")
+
+            # 异步查询
+            is_running = True
+            while is_running:
+                # url_query = f'https://dashscope.aliyuncs.com/api/v1/tasks/{task_id}'
+                url_query = f'https://poc-dashscope.aliyuncs.com/api/v1/tasks/{task_id}'
+                try:
+                    res_ = requests.post(url_query, headers=headers, timeout=60)
+                except requests.Timeout:
+                    # back off and retry
+                    raise gr.Error("网络波动，请求失败，请再次尝试")
+                respose_code = res_.status_code
+                if 200 == respose_code:
+                    res = json.loads(res_.content.decode())
+                    if "SUCCEEDED" == res['output']['task_status']:
+                        logger.info(f"task_id: {task_id}: Generation task query success.")
+                        results = res['output']
+                        img_urls = results['output_img']
+                        logger.info(f"task_id: {task_id}: {res}")
+                        break
+                    elif "FAILED" != res['output']['task_status']:
+                        logger.debug(f"task_id: {task_id}: query result...")
+                        time.sleep(1)
+                    else:
+                        raise gr.Error('Fail to get results from Generation task.')
+
+                else:
+                    logger.error(f'task_id: {task_id}: Fail to query task result: {res_.content}')
+                    raise gr.Error("Fail to query task result.")
+
+            logger.info(f"task_id: {task_id}: download generated images.")
+            img_data = download_svgs(img_urls, BATCH_SIZE, svg_name)
+            logger.info(f"task_id: {task_id}: Generate done.")
+            all_image_data += img_data
+        else:
+            logger.error(f'Fail to create Generation task: {res_.content}')
+            raise gr.Error("Fail to create Generation task.")
+
+    if len(all_image_data) != REPEAT * BATCH_SIZE:
+        raise gr.Error("Fail to Generation.")
+    return all_image_data[-1:]
+
+if __name__ == "__main__":
+    call_generation()

demo/instructions.py

@@ -0,0 +1,31 @@
+INSTRUCTIONS_VECTORIZE_SIMPLIFY='''
+<style>
+        body {
+            font-family: Arial, sans-serif;
+            line-height: 1.6;
+            margin: 20px;
+        }
+        ul {
+            list-style-type: disc;
+            margin-left: 20px;
+        }
+    </style>
+
+<body>
+
+<p>Step1: 上传一个需要矢量化的图片 Upload a image to vectorize</p>
+
+<p>Step2: 调整矢量化参数 Adjust the parameters</p>
+<ul>
+    <li>使用 GPU 加速 (GPU Accelerate): 对于颜色复杂的图形, 建议使用GPU加速</li>
+    <li>超分倍数 (Super Resolution Scale): suggested for small image 对于小图片, 可以选择 2-4 倍, 可提升矢量化效果</li>
+    <li>简化矢量图 (Simplify SVG): reduce control points number 是否简化矢量图, 可减少控制点数量</li>
+    <li>优化矢量图 (Optimize SVG): optimize svg to make it closer to input 是否优化矢量图, 可使得矢量图更接近原图,推荐配合cutout模式使用,可减少图形之间空隙</li>
+    <li>小面积图形 (Small Speckles): remove small areas 是否移除小面积图形</li>
+    <li>排序方法 (Sorting Method): sort shape in lightness or area 色块排序方式, 可以选择按亮度排序或者面积排序</li>
+    <li>排序顺序 (Sorting Order): sort in ascend or descend 从大到小或者从小到大</li>
+    <li>追踪模式 (Trace Mode): trace in cutout or overlap mode 路径追踪模式, 可以选择 Cutout 或者 Overlap。Cutout 模式图形没有重叠,Overlap 模式图形会根据排序方法和顺序进行重叠</li>
+</ul>
+<p>Step3: 点击转化矢量图 Press the button to vectorize</p>
+</body>
+'''

demo/log.py

@@ -0,0 +1,18 @@
+import logging
+import os
+from logging.handlers import RotatingFileHandler
+
+log_file_name = "workdir/log_vectorize_anything.log"
+os.makedirs(os.path.dirname(log_file_name), exist_ok=True)
+
+format = '[%(levelname)s] %(asctime)s "%(filename)s", line %(lineno)d, %(message)s'
+logging.basicConfig(
+    format=format,
+    datefmt="%Y-%m-%d %H:%M:%S",
+    level=logging.INFO)
+logger = logging.getLogger(name="WordArt_Studio")
+
+fh = RotatingFileHandler(log_file_name, maxBytes=20000000, backupCount=3)
+formatter = logging.Formatter(format, datefmt="%Y-%m-%d %H:%M:%S")
+fh.setFormatter(formatter)
+logger.addHandler(fh)

demo/util.py

@@ -0,0 +1,104 @@
+import concurrent.futures
+import io
+import os
+
+import numpy as np
+import oss2
+import requests
+from PIL import Image, ImageDraw, ImageFont
+
+from .log import logger
+
+# oss
+access_key_id = os.getenv("ACCESS_KEY_ID")
+access_key_secret = os.getenv("ACCESS_KEY_SECRET")
+bucket_name = os.getenv("BUCKET_NAME")
+endpoint = os.getenv("ENDPOINT")
+
+oss_path = "VectorizeAnything"
+oss_path_img_gallery = "VectorizeAnything"
+
+bucket = oss2.Bucket(oss2.Auth(access_key_id, access_key_secret), endpoint, bucket_name)
+
+def download_img_pil(index, img_url):
+    # print(img_url)
+    r = requests.get(img_url, stream=True)
+    if r.status_code == 200:
+        img = Image.open(io.BytesIO(r.content))
+        return (index, img)
+    else:
+        logger.error(f"Fail to download: {img_url}")
+
+
+def download_images(img_urls, batch_size):
+    imgs_pil = [None] * batch_size
+    # worker_results = []
+    with concurrent.futures.ThreadPoolExecutor(max_workers=4) as executor:
+        to_do = []
+        for i, url in enumerate(img_urls):
+            future = executor.submit(download_img_pil, i, url)
+            to_do.append(future)
+
+        for future in concurrent.futures.as_completed(to_do):
+            ret = future.result()
+            # worker_results.append(ret)
+            index, img_pil = ret
+            imgs_pil[index] = img_pil  # 按顺序排列url，后续下载关联的图片或者svg需要使用
+
+    return imgs_pil
+
+def download_svg(index, img_url, svg_name):
+    # print(img_url)
+    r = requests.get(img_url, stream=True)
+    if r.status_code == 200:
+        img = io.BytesIO(r.content)
+        svg_url = upload_svg_2_oss(img, f"{svg_name}_{index}.svg")
+        return (index, svg_url)
+    else:
+        logger.error(f"Fail to download: {img_url}")
+
+
+def download_svgs(img_urls, batch_size, svg_name):
+    imgs_pil = [None] * batch_size
+    # worker_results = []
+    with concurrent.futures.ThreadPoolExecutor(max_workers=4) as executor:
+        to_do = []
+        for i, url in enumerate(img_urls):
+            future = executor.submit(download_svg, i, url, svg_name)
+            to_do.append(future)
+
+        for future in concurrent.futures.as_completed(to_do):
+            ret = future.result()
+            # worker_results.append(ret)
+            index, img_pil = ret
+            imgs_pil[index] = img_pil  # 按顺序排列url，后续下载关联的图片或者svg需要使用
+
+    return imgs_pil
+
+def upload_np_2_oss(input_image, name="cache.png", gallery=False):
+    imgByteArr = io.BytesIO()
+    Image.fromarray(input_image).save(imgByteArr, format="PNG")
+    imgByteArr = imgByteArr.getvalue()
+
+    if gallery:
+        path = oss_path_img_gallery
+    else:
+        path = oss_path
+
+    bucket.put_object(path + "/" + name, imgByteArr)  # data为数据，可以是图片
+    ret = bucket.sign_url('GET', path + "/" + name, 60 * 60 * 24)  # 返回值为链接，参数依次为，方法/oss上文件路径/过期时间(s)
+    del imgByteArr
+    return ret
+
+
+def upload_svg_2_oss(input_svg, name="cache.png", gallery=False):
+
+    if gallery:
+        path = oss_path_img_gallery
+    else:
+        path = oss_path
+
+    bucket.put_object(path + "/" + name, input_svg)  # data为数据，可以是图片
+    ret = bucket.sign_url('GET', path + "/" + name, 60 * 60 * 24)  # 返回值为链接，参数依次为，方法/oss上文件路径/过期时间(s)
+    del input_svg
+    return ret