Can't install on Windows 10 + Anaconda - Running = Errors ??

[AlonDan]

I've created an environment called: "rembg-greenscreen" with python 3.9 and activated it
I tried to install using: pip install rembg-greenscreen

But I got some errors I'm not sure how to fix as I'm not a programmer,

Can someone please explain how to install it correct without errors?

Thanks ahead!

[AlonDan]

Problem solved by manually install numpy:
pip install numpy==1.19.4

Then, install again should complete without any errors:
pip install rembg-greenscreen

I hope this will help others in case you ran in to this issue. 👍

[AlonDan]

I keep on update in case anyone can help, or if I'll find a solution.

I've tried to run my first test, it seems like it's "almost" working... but I get errors and it's stuck.
I can get back to Anaconda using Ctrl+C

I'm not sure what to do now or how to fix it... any help will be appreciated.
Thanks ahead!

[hic9507]

Did you solve this problem?
I have an same error.
If you solve it, I need your help.
Thanks

[AlonDan]

Did you solve this problem? I have an same error. If you solve it, I need your help. Thanks

Unfortunately I couldn't solve it,
I even tried to reinstall from scratch but got the same errors.

[hic9507]

I saw other issue on 'rembg'.
[(https://github.com/danielgatis/rembg/issues?q=pickle)]

I tried download 'u2net_human_seg.pth', and move it right place.

But, this file( u2net_human_seg.pth) keeps downloading, and after only 3kb is downloaded, the same error is shown.

So I'm trying to find out if there is a part that makes the model download unconditionally, and I'm curious if it can be solved after seeing this.
(rembg-screen/src/rembg/u2net/bg.py)
(rembg-screen/src/rembg/u2net/multiprocessing.py)

I am a newb, so I think you are better than me for solve this problem.

Anyway, I will refer to the link, so please read the problem carefully.
They had a similar problem.
And if any of us find a solution first, please comment here.

[hic9507]

Did you solve this problem? I have an same error. If you solve it, I need your help. Thanks

Unfortunately I couldn't solve it, I even tried to reinstall from scratch but got the same errors.

Did you solve this problem?

[AlonDan]

And if any of us find a solution first, please comment here.

I also had this problem, where it download to my C drive under the user and some directory path (I can't remember) and when trying to run it OVERWRITE the u2net_human_seg.pth (172MB size) with another one and stuck... probably the 3KB like you show.

I tried to copy it MANUALLY but it will overwrite the moment I try to run via anaconda again.

Can you please update you link? it leads to the main issues page
I would like to follow your suggestion and read it maybe I'll give it a try.
I'm a newbie in this and not a programmer so if it's complex I'll probably won't be able to do much.

Like you suggested @hic9507 if any of us find a solution, let's share it here ❤

[suffa07]

I experience a similar problem when attempting to install "rembg-greenscreen", repeatedly received an error: "could not build numpy wheel...". I tried a number of things, but ultimately what seemed to work was running the following commands (in this order) from the cmd line (OSX 13.2.1):

1. sudo pip install --upgrade --no-deps --force-reinstall numpy
2. sudo pip install --upgrade --no-deps --force-reinstall rembg-greenscreen

[maximedrn]

Hi, here's a quick solution if anyone encounters this problem.

• Download the u2net_human_seg.pth weight file here and place it in the <user_directory>/.u2net/ folder.
• Delete lines 64 to 73 of the rembg/bg.py file. You'll have the correct weight file and it won't be overridden.

import os
import typing

import moviepy.editor as mpy
import numpy as np
import requests
import torch
import torch.nn.functional
import torch.nn.functional
from hsh.library.hash import Hasher
from tqdm import tqdm

from .u2net import u2net

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


def iter_frames(path):
    return mpy.VideoFileClip(path).resize(height=320).iter_frames(dtype="uint8")


class Net(torch.nn.Module):
    def __init__(self, model_name):
        super(Net, self).__init__()
        hasher = Hasher()

        model, hash_val, drive_target, env_var = {
            'u2netp':          (u2net.U2NETP,
                                'e4f636406ca4e2af789941e7f139ee2e',
                                '1rbSTGKAE-MTxBYHd-51l2hMOQPT_7EPy',
                                'U2NET_PATH'),
            'u2net':           (u2net.U2NET,
                                '09fb4e49b7f785c9f855baf94916840a',
                                '1-Yg0cxgrNhHP-016FPdp902BR-kSsA4P',
                                'U2NET_PATH'),
            'u2net_human_seg': (u2net.U2NET,
                                '347c3d51b01528e5c6c071e3cff1cb55',
                                '1ao1ovG1Qtx4b7EoskHXmi2E9rp5CHLcZ',
                                'U2NET_PATH')
            }[model_name]
        path = os.environ.get(env_var, os.path.expanduser(os.path.join("~", ".u2net", model_name + ".pth")))
        net = model(3, 1)
        if not os.path.exists(path) or hasher.md5(path) != hash_val:
            head, tail = os.path.split(path)
            os.makedirs(head, exist_ok=True)

            URL = "https://docs.google.com/uc?export=download"

            session = requests.Session()
            response = session.get(URL, params={"id": drive_target}, stream=True)

            token = None
            for key, value in response.cookies.items():
                if key.startswith("download_warning"):
                    token = value
                    break

            if token:
                params = {"id": drive_target, "confirm": token}
                response = session.get(URL, params=params, stream=True)

            total = int(response.headers.get("content-length", 0))

            # with open(path, "wb") as file, tqdm(
            #     desc=f"Downloading {tail} to {head}",
            #     total=total,
            #     unit="iB",
            #     unit_scale=True,
            #     unit_divisor=1024,
            #     ) as bar:
            #     for data in response.iter_content(chunk_size=1024):
            #         size = file.write(data)
            #         bar.update(size)
        net.load_state_dict(torch.load(path, map_location=torch.device(DEVICE)))
        net.to(device=DEVICE, dtype=torch.float32, non_blocking=True)
        net.eval()
        self.net = net

    def forward(self, block_input: torch.Tensor):
        image_data = block_input.permute(0, 3, 1, 2)
        original_shape = image_data.shape[2:]
        image_data = torch.nn.functional.interpolate(image_data, (320, 320), mode='bilinear')
        image_data = (image_data / 255 - 0.485) / 0.229
        out = self.net(image_data)[:, 0:1]
        ma = torch.max(out)
        mi = torch.min(out)
        out = (out - mi) / (ma - mi) * 255
        out = torch.nn.functional.interpolate(out, original_shape, mode='bilinear')
        out = out[:, 0]
        out = out.to(dtype=torch.uint8, device=torch.device('cpu'), non_blocking=True).detach()
        return out


@torch.no_grad()
def remove_many(image_data: typing.List[np.array], net: Net):
    image_data = np.stack(image_data)
    image_data = torch.as_tensor(image_data, dtype=torch.float32, device=DEVICE)
    return net(image_data).numpy()