Model Detectron2 di Pytorch

By | July 30, 2024
0 Comment
Print Friendly, PDF & Email
222 Views

Detectron2 adalah library open-source yang dikembangkan oleh Facebook AI Research (FAIR) untuk melakukan deteksi objek. Library ini merupakan penerus dari Detectron dan dirancang untuk menjadi lebih modular, fleksibel, dan efisien dalam berbagai aplikasi visi komputer, seperti deteksi objek, segmentasi instance, dan segmentasi semantik. Berikut beberapa fitur dan penjelasan tentang Detectron2:

  1. Modular dan Fleksibel:
    • Detectron2 memiliki arsitektur yang modular, sehingga komponen-komponen seperti backbone, kepala (head), dan loss function dapat dengan mudah dikombinasikan dan dikustomisasi sesuai kebutuhan.
    • Library ini mendukung berbagai model backbone seperti ResNet, ResNeXt, dan FPN.
  2. Kinerja Tinggi:
    • Optimisasi yang dilakukan pada Detectron2 membuatnya sangat efisien dalam hal penggunaan memori dan waktu komputasi.
    • Dukungan penuh untuk pemrosesan paralel dengan GPU memungkinkan eksekusi yang cepat dan efisien.
  3. Implementasi SOTA (State-Of-The-Art):
    • Detectron2 menyediakan implementasi dari berbagai algoritma deteksi dan segmentasi objek terkini yang sering digunakan dalam penelitian dan industri.
    • Contoh model yang didukung termasuk Faster R-CNN, Mask R-CNN, RetinaNet, dan lain-lain.
  4. Dukungan untuk Pelatihan dan Inference:
    • Detectron2 mendukung pelatihan model dari awal (scratch) atau fine-tuning model yang sudah dilatih sebelumnya.
    • Library ini juga mendukung inference atau prediksi dengan model yang sudah dilatih.
  5. Integrasi yang Mudah:
    • Detectron2 dibangun di atas PyTorch, membuatnya mudah diintegrasikan dengan ekosistem PyTorch dan framework deep learning lainnya.
    • API yang disediakan user-friendly dan dirancang untuk memudahkan peneliti dan praktisi dalam mengembangkan dan menguji model.
  6. Dokumentasi dan Komunitas:
    • Detectron2 memiliki dokumentasi yang komprehensif serta banyak contoh dan tutorial untuk membantu pengguna memulai.
    • Komunitas yang aktif dan dukungan dari Facebook AI Research memastikan bahwa library ini terus diperbarui dan diperbaiki.

Cara Install Detectron2

Kalian bisa mengikuti langkah berikut

https://colab.research.google.com/drive/16jcaJoc6bCFAQ96jDe2HwtXj7BMD_-m5

Untuk detectron2 saat ini masih menggunakan torch version 2 kebawah

https://detectron2.readthedocs.io/en/latest/tutorials/install.html

It would be useful to have the wheels for newer versions of torch. It seems like they haven’t been updated since November 2021. I am using the more limited installation method for now:

mari kita coba menggunakan perintah

!python -m pip install 'git+https://github.com/facebookresearch/detectron2.git@v0.6'

tetap nggak bisa! karena emang terakhir diperbaharui yaitu 2011

https://github.com/facebookresearch/detectron2/releases

Sehingga kita harus menggunakan python 3.7 untuk bisa install torch 1.9.0

Membuat virtual environment pada conda dengan version python 3.8

Kalian bisa menggunakan cara berikut, misalkan kita akan membuat virtual environment dengan nama virtual1

Minimal untuk menggunakan versi 3.8 karena ada error jika menggunakan 3.7

ImportError: cannot import name ‘cached_property’ from ‘functools’ (/opt/anaconda3/envs/virtual1/lib/python3.7/functools.py)

conda create -n virtual1 python==3.8

kemudian pilih yes saja, seperti berikut tampilannya

(base) user@MacBook-Air detectron2 % conda create -n virtual1 python==3.8
Channels:
 - defaults
Platform: osx-64
Collecting package metadata (repodata.json): done
Solving environment: done

## Package Plan ##

  environment location: /opt/anaconda3/envs/virtual1

  added / updated specs:
    - python==3.7


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    ca-certificates-2024.7.2   |       hecd8cb5_0         128 KB
    certifi-2022.12.7          |   py37hecd8cb5_0         151 KB
    libffi-3.2.1               |    h0a44026_1007          43 KB
    openssl-1.0.2u             |       h1de35cc_0         2.0 MB
    pip-22.3.1                 |   py37hecd8cb5_0         2.7 MB
    python-3.7.0               |       hc167b69_0        13.2 MB
    readline-7.0               |       h1de35cc_5         316 KB
    setuptools-65.6.3          |   py37hecd8cb5_0         1.1 MB
    sqlite-3.33.0              |       hffcf06c_0         1.3 MB
    tk-8.6.14                  |       h4d00af3_0         3.4 MB
    wheel-0.38.4               |   py37hecd8cb5_0          65 KB
    xz-5.4.6                   |       h6c40b1e_1         371 KB
    zlib-1.2.13                |       h4b97444_1         102 KB
    ------------------------------------------------------------
                                           Total:        24.8 MB

The following NEW packages will be INSTALLED:

  ca-certificates    pkgs/main/osx-64::ca-certificates-2024.7.2-hecd8cb5_0 
  certifi            pkgs/main/osx-64::certifi-2022.12.7-py37hecd8cb5_0 
  libcxx             pkgs/main/osx-64::libcxx-14.0.6-h9765a3e_0 
  libedit            pkgs/main/osx-64::libedit-3.1.20230828-h6c40b1e_0 
  libffi             pkgs/main/osx-64::libffi-3.2.1-h0a44026_1007 
  ncurses            pkgs/main/osx-64::ncurses-6.4-hcec6c5f_0 
  openssl            pkgs/main/osx-64::openssl-1.0.2u-h1de35cc_0 
  pip                pkgs/main/osx-64::pip-22.3.1-py37hecd8cb5_0 
  python             pkgs/main/osx-64::python-3.7.0-hc167b69_0 
  readline           pkgs/main/osx-64::readline-7.0-h1de35cc_5 
  setuptools         pkgs/main/osx-64::setuptools-65.6.3-py37hecd8cb5_0 
  sqlite             pkgs/main/osx-64::sqlite-3.33.0-hffcf06c_0 
  tk                 pkgs/main/osx-64::tk-8.6.14-h4d00af3_0 
  wheel              pkgs/main/osx-64::wheel-0.38.4-py37hecd8cb5_0 
  xz                 pkgs/main/osx-64::xz-5.4.6-h6c40b1e_1 
  zlib               pkgs/main/osx-64::zlib-1.2.13-h4b97444_1 


Proceed ([y]/n)? y


Downloading and Extracting Packages:
                                                                                                                                                                                  
Preparing transaction: done                                                                                                                                                       
Verifying transaction: done                                                                                                                                                       
Executing transaction: done                                                                                                                                                       
#                                                                                                                                                                                 
# To activate this environment, use                                                                                                                                               
#                                                                                                                                                                                 
#     $ conda activate virtual1                                                                                                                                                   
#                                                                                                                                                                                 
# To deactivate an active environment, use                                                                                                                                        
#                                                                                                                                                                                 
#     $ conda deactivate

Activate Virtual Environment

seperti biasa, kita install dengan perintah conda activate virtual1

(base) user@MacBook-Air detectron2 % conda activate virtual1 
(virtual1) user@MacBook-Air detectron2 % python --version
Python 3.7.0

Install torch 1.9

Lanjut untuk install torch versi 1.9

(virtual1) user@MacBook-Air detectron2 % pip install torch==1.9
Collecting torch==1.9
  Downloading torch-1.9.0-cp37-none-macosx_10_9_x86_64.whl (127.9 MB)
     ━━━━━━━━━━━━━━━━━━━━╺━━━━━━━━━━━━━━━━━━━ 64.7/127.9 MB 186.8 kB/s eta 0:05:39

Install Opencv

jangan lupa install opencv menggunakan version 3.4.10.37

pip install opencv-python==3.4.10.37

Install TorchVision

dan juga torchvision dengan version 0.10.0

pip install torchvision==0.10.0

Install detectron2

jangan lupa kalian pengguna mac, buat perintah berikut di terminal

See also  Belajar RNN untuk kasus Sederhana

https://stackoverflow.com/questions/56083725/macos-build-issues-lstdc-not-found-while-building-python-package

yaitu dengan memberikan perintah

export MACOSX_DEPLOYMENT_TARGET=10.9

 

setelah itu, dilanjut dengan install detectron2

python -m pip install 'git+https://github.com/facebookresearch/detectron2.git'

Mencoba Detectron2

Mari kita test import package nya, apakah berhasil?

from detectron2 import model_zoo
from detectron2.engine import DefaultPredictor
from detectron2.config import get_cfg
from detectron2.utils.visualizer import Visualizer
from detectron2.data import MetadataCatalog, DatasetCatalog

kalau error pada saat install

pip install cython

juga install

pip install pyyaml

kalau masih ada problem bisa baca https://github.com/facebookresearch/detectron2/issues/5216

Berikut hasil running detectron2 di Linux

Saya menggunakan VPS linux dengan informasi sebagai berikut

(vm2) root@upbeat-turing:/home/bejo# screenfetch
                          ./+o+-       root@upbeat-turing.xxx-xx-xx-xx.plesk.page
                  yyyyy- -yyyyyy+      OS: Ubuntu 22.04 jammy
               ://+//////-yyyyyyo      Kernel: x86_64 Linux 5.15.0-46-generic
           .++ .:/++++++/-.+sss/`      Uptime: 10d 17h 29m
         .:++o:  /++++++++/:--:/-      Packages: 1292
        o:+o+:++.`..```.-/oo+++++/     Shell: bash 5.1.16
       .:+o:+o/.          `+sssoo+/    Disk: 48G / 158G (32%)
  .++/+:+oo+o:`             /sssooo.   CPU: Intel Xeon E5-2680 v3 @ 4x 2.5GHz
 /+++//+:`oo+o               /::--:.   GPU: Cirrus Logic GD 5446
 \+/+o+++`o++o               ++////.   RAM: 7717MiB / 7949MiB
  .++.o+++oo+:`             /dddhhh.  
       .+.o+oo:.          `oddhhhh+   
        \+.++o+o``-````.:ohdhhhhh+    
         `:o+++ `ohhhhhhhhyo++os:     
           .o:`.syhhhhhhh/.oo++o`     
               /osyyyyyyo++ooo+++/    
                   ````` +oo+++o\:    
                          `oo++.

yang sebelumnya menggunakan vm dengan informasi sebagai berikut

Python 3.8.0 (default, Nov  6 2019, 21:49:08) 
[GCC 7.3.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>>

versi torch, torchvision, dan opencv sebagai berikut

>>> import torch
>>> import torchvision
>>> import cv2
>>> torch.__version__
'1.9.0+cu102'
>>> torchvision.__version__
'0.10.0+cu102'
>>> cv2.__version__
'4.10.0'
>>>

Bila kalian punya GPU wajib install

apt install nvidia-cuda-toolkit --fix-missing

jika tidak punya GPU wajib setting CPU, mari kita coba

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Thu Jul 25 14:03:04 2024

@author: user
"""

import sys, os, distutils.core
import torch, detectron2


TORCH_VERSION = ".".join(torch.__version__.split(".")[:2])
CUDA_VERSION = torch.__version__.split("+")[-1]
print("torch: ", TORCH_VERSION, "; cuda: ", CUDA_VERSION)
print("detectron2:", detectron2.__version__)


import detectron2
from detectron2.utils.logger import setup_logger
setup_logger()

# import some common libraries
import numpy as np
import os, json, cv2, random
# from google.colab.patches import cv2_imshow

# import some common detectron2 utilities
from detectron2 import model_zoo
from detectron2.engine import DefaultPredictor
from detectron2.config import get_cfg
from detectron2.utils.visualizer import Visualizer
from detectron2.data import MetadataCatalog, DatasetCatalog

im = cv2.imread("image.jpeg")
cfg = get_cfg()
# add project-specific config (e.g., TensorMask) here if you're not running a model in detectron2's core library
cfg.merge_from_file(model_zoo.get_config_file("COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml"))
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5  # set threshold for this model
# Find a model from detectron2's model zoo. You can use the https://dl.fbaipublicfiles... url as well
cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml")
cfg.MODEL.DEVICE = 'cpu'
predictor = DefaultPredictor(cfg)
outputs = predictor(im)

karena menggunakan CPU tentu hasilnya akan sangat lama sekali! berikut hasil uji coba diatas

(vm2) root@upbeat-turing:/home/bejo# python
Python 3.8.0 (default, Nov  6 2019, 21:49:08) 
[GCC 7.3.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys, os, distutils.core
>>> import torch, detectron2
>>> 
>>> 
>>> TORCH_VERSION = ".".join(torch.__version__.split(".")[:2])
>>> CUDA_VERSION = torch.__version__.split("+")[-1]
>>> print("torch: ", TORCH_VERSION, "; cuda: ", CUDA_VERSION)
torch:  1.9 ; cuda:  cu102
>>> print("detectron2:", detectron2.__version__)
detectron2: 0.6
>>> 
>>> 
>>> import detectron2
>>> from detectron2.utils.logger import setup_logger
>>> setup_logger()
<Logger detectron2 (DEBUG)>
>>> 
>>> # import some common libraries
... import numpy as np
>>> import os, json, cv2, random
>>> # from google.colab.patches import cv2_imshow
... 
>>> # import some common detectron2 utilities
... from detectron2 import model_zoo
>>> from detectron2.engine import DefaultPredictor
>>> from detectron2.config import get_cfg
>>> from detectron2.utils.visualizer import Visualizer
>>> from detectron2.data import MetadataCatalog, DatasetCatalog
>>> im = cv2.imread("image.jpeg")
>>> cfg = get_cfg()
>>> # add project-specific config (e.g., TensorMask) here if you're not running a model in detectron2's core library
... cfg.merge_from_file(model_zoo.get_config_file("COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml"))
>>> cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5  # set threshold for this model
>>> # Find a model from detectron2's model zoo. You can use the https://dl.fbaipublicfiles... url as well
... cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml")
>>> cfg.MODEL.DEVICE = 'cpu'
>>> predictor = DefaultPredictor(cfg)
[07/27 23:21:54 d2.checkpoint.detection_checkpoint]: [DetectionCheckpointer] Loading from https://dl.fbaipublicfiles.com/detectron2/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x/137849600/model_final_f10217.pkl ...
model_final_f10217.pkl: 178MB [00:40, 4.36MB/s]                                                                                                                                   
>>> outputs = predictor(im)
/root/anaconda3/envs/vm2/lib/python3.8/site-packages/torch/nn/functional.py:718: UserWarning: Named tensors and all their associated APIs are an experimental feature and subject to change. Please do not use them for anything important until they are released as stable. (Triggered internally at  /pytorch/c10/core/TensorImpl.h:1156.)
  return torch.max_pool2d(input, kernel_size, stride, padding, dilation, ceil_mode)
>>> outputs
{'instances': Instances(num_instances=15, image_height=480, image_width=640, fields=[pred_boxes: Boxes(tensor([[126.6035, 244.8977, 459.8291, 480.0000],
        [251.1083, 157.8127, 338.9731, 413.6379],
        [114.8496, 268.6864, 148.2351, 398.8111],
        [  0.8217, 281.0327,  78.6072, 478.4209],
        [ 49.3953, 274.1229,  80.1545, 342.9808],
        [561.2248, 271.5816, 596.2755, 385.2552],
        [385.9072, 270.3125, 413.7130, 304.0397],
        [515.9295, 278.3743, 562.2792, 389.3803],
        [335.2410, 251.9167, 414.7491, 275.9375],
        [350.9300, 269.2060, 386.0984, 297.9081],
        [331.6292, 230.9996, 393.2759, 257.2009],
        [510.7349, 263.2656, 570.9865, 295.9194],
        [409.0841, 271.8646, 460.5582, 356.8722],
        [506.8767, 283.3257, 529.9403, 324.0392],
        [594.5663, 283.4820, 609.0577, 311.4124]])), scores: tensor([0.9997, 0.9957, 0.9915, 0.9882, 0.9861, 0.9840, 0.9769, 0.9716, 0.9062,
        0.9037, 0.8870, 0.8575, 0.6592, 0.5899, 0.5767]), pred_classes: tensor([17,  0,  0,  0,  0,  0,  0,  0, 25,  0, 25, 25,  0,  0, 24]), pred_masks: tensor([[[False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         ...,
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False]],

        [[False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         ...,
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False]],

        [[False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         ...,
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False]],

        ...,

        [[False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         ...,
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False]],

        [[False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         ...,
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False]],

        [[False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         ...,
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False],
         [False, False, False,  ..., False, False, False]]])])}

lebih lanjut ref: https://colab.research.google.com/drive/16jcaJoc6bCFAQ96jDe2HwtXj7BMD_-m5

See also  Akhirnya Torch versi R telah tersedia - Framework Machine Learning

Untu lebih jelasnya mengenai inference detectron2 bisa dibaca link diatas

untuk menampilkan hasilnya kedalam gambar yaitu

from matplotlib import pyplot as plt


v = Visualizer(im[:, :, ::-1], MetadataCatalog.get(cfg.DATASETS.TRAIN[0]), scale=1.2)
out = v.draw_instance_predictions(outputs["instances"].to("cpu"))



plt.figure, plt.imshow(out.get_image()[:, :, ::-1]),plt.show()

 

Berikut hasil install detectron2 Mac OS

dengan spesifikasi sebagai berikut

(base) user@MacBook-Air / % screenfetch                    
                               
                 -/+:.          user@MacBook-Air
                :++++.          OS: 64bit macOS  
               /+++/.           Kernel: x86_64 Darwin 21.6.0
       .:-::- .+/:-``.::-       Uptime: 6h 15m
    .:/++++++/::::/++++++/:`    Packages: 146
  .:///////////////////////:`   Shell: zsh 5.8.1
  ////////////////////////`     Resolution: 1440x900 
 -+++++++++++++++++++++++`      DE: Aqua
 /++++++++++++++++++++++/       WM: Quartz Compositor
 /sssssssssssssssssssssss.      WM Theme: Blue
 :ssssssssssssssssssssssss-     Font: SFMonoRegular
  osssssssssssssssssssssssso/`  Disk: 111G / 251G (48%)
  `syyyyyyyyyyyyyyyyyyyyyyyy+`  CPU: Intel Core i5-5250U @ 1.60GHz
   `ossssssssssssssssssssss/    GPU: Intel HD Graphics 6000 
     :ooooooooooooooooooo+.     RAM: 3353MiB / 8192MiB
      `:+oo+/:-..-:/+o+/-

 

import sys, os, distutils.core
import torch, detectron2


TORCH_VERSION = ".".join(torch.__version__.split(".")[:2])
CUDA_VERSION = torch.__version__.split("+")[-1]
print("torch: ", TORCH_VERSION, "; cuda: ", CUDA_VERSION)
print("detectron2:", detectron2.__version__)
torch:  1.9 ; cuda:  1.9.0
detectron2: 0.6