Detectron2 Custom Object Detection + Sahi

[drdoof2019]

I am using colab. I trained detectron2 for custom object detection (1 class named "hata") and got "best.pth" file in my google drive. I mounted drive and used the code below:

!pip install -U sahi
!pip install torch==2.0.0
!pip install torchvision==0.15.1
!pip install opencv-python==4.6.0.66
!pip install matplotlib==3.5.3
!pip install git+https://github.com/facebookresearch/detectron2.git
#pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cpu/torch1.10/index.html # for Detectron2-cpu
#pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu111/torch1.10/index.html # for Detectron2-cuda11.1

# will be used for detectron2 fasterrcnn model zoo name
from sahi.utils.detectron2 import Detectron2TestConstants

# import required functions, classes
from sahi import AutoDetectionModel
from sahi.predict import get_sliced_prediction, predict, get_prediction
from sahi.utils.file import download_from_url
from sahi.utils.cv import read_image
from IPython.display import Image


# set detectron2 fasterrcnn model zoo name
model_path = Detectron2TestConstants.FASTERCNN_MODEL_ZOO_NAME
model_path = "/content/drive/MyDrive/Datasets/train-object-detector-detectron2/output/best.pth"


# download test images into demo_data folder
download_from_url('https://raw.githubusercontent.com/obss/sahi/main/demo/demo_data/small-vehicles1.jpeg', 'demo_data/small-vehicles1.jpeg')
download_from_url('https://raw.githubusercontent.com/obss/sahi/main/demo/demo_data/terrain2.png', 'demo_data/terrain2.png')


detection_model = AutoDetectionModel.from_pretrained(
    model_type='detectron2',
    model_path=model_path,
    config_path=model_path,
    confidence_threshold=0.5,
    image_size=640,
    device="cpu", # or 'cuda:0'
)

model_path = Detectron2TestConstants.FASTERCNN_MODEL_ZOO_NAME works but when I use model_path = "/content/drive/MyDrive/Datasets/train-object-detector-detectron2/output/best.pth" error below occured:

/content/drive/MyDrive/Datasets/train-object-detector-detectron2/output/best.pth not available in Model Zoo!
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
[/usr/local/lib/python3.10/dist-packages/sahi/models/detectron2.py](https://localhost:8080/#) in load_model(self)
     29         try:  # try to load from model zoo
---> 30             config_file = model_zoo.get_config_file(self.config_path)
     31             cfg.merge_from_file(config_file)

13 frames
RuntimeError: /content/drive/MyDrive/Datasets/train-object-detector-detectron2/output/best.pth not available in Model Zoo!

During handling of the above exception, another exception occurred:

UnicodeDecodeError                        Traceback (most recent call last)
[/usr/lib/python3.10/codecs.py](https://localhost:8080/#) in decode(self, input, final)
    320         # decode input (taking the buffer into account)
    321         data = self.buffer + input
--> 322         (result, consumed) = self._buffer_decode(data, self.errors, final)
    323         # keep undecoded input until the next call
    324         self.buffer = data[consumed:]

UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 64: invalid start byte

[HauCongNguyen]

Did you solve it ? I had the answer.

[drdoof2019]

No I couldn't use Detectron2 custom object detectron with SAHI

[HauCongNguyen]

You can add the following code to the detectron2 config, and you will get config.yaml file. The config_path is the path to config.yaml :
from sahi.utils.detectron2 import export_cfg_as_yaml
export_cfg_as_yaml(cfg, export_path='config.yaml')

---

Ex:

cfg = get_cfg()

cfg.merge_from_file(model_zoo.get_config_file("COCO-Detection/faster_rcnn_R_50_FPN_3x.yaml"))
cfg.DATASETS.TRAIN = ("",)
cfg.DATASETS.TEST = ("",)
cfg.DATALOADER.NUM_WORKERS = 2
.............
cfg.SOLVER.GAMMA = 0.05
cfg.MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE = 64
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 1

cfg.TEST.EVAL_PERIOD=500

#export config.yaml
from sahi.utils.detectron2 import export_cfg_as_yaml
export_cfg_as_yaml(cfg, export_path='config.yaml')
--------------------------------

[hkeay]

I've also tried this but couldn't get it to work, what do you use for the model path?

[HauCongNguyen]

Model path is the path to your trained weight.