SerdarHelli's picture
Update app.py
ce7d026
raw
history blame
4.6 kB
import gradio as gr
import sahi.utils
from sahi import AutoDetectionModel
import sahi.predict
import sahi.slicing
from PIL import Image
import numpy
from huggingface_hub import hf_hub_download
import torch
IMAGE_SIZE = 640
model_path=hf_hub_download("kadirnar/deprem_model_v1", filename="last.pt",revision="main")
current_device='cuda' if torch.cuda.is_available() else 'cpu'
# Model
model = AutoDetectionModel.from_pretrained(
model_type="yolov5", model_path=model_path, device=current_device, confidence_threshold=0.5, image_size=IMAGE_SIZE
)
def sahi_yolo_inference(
model_type,
image,
slice_height=512,
slice_width=512,
overlap_height_ratio=0.2,
overlap_width_ratio=0.2,
postprocess_type="GREEDYNMM",
postprocess_match_metric="IOS",
postprocess_match_threshold=0.5,
postprocess_class_agnostic=False,
):
#image_width, image_height = image.size
# sliced_bboxes = sahi.slicing.get_slice_bboxes(
# image_height,
# image_width,
# slice_height,
# slice_width,
# False,
# overlap_height_ratio,
# overlap_width_ratio,
# )
# if len(sliced_bboxes) > 60:
# raise ValueError(
# f"{len(sliced_bboxes)} slices are too much for huggingface spaces, try smaller slice size."
# )
if "SAHI" in model_type:
prediction_result_2 = sahi.predict.get_sliced_prediction(
image=image,
detection_model=model,
slice_height=int(slice_height),
slice_width=int(slice_width),
overlap_height_ratio=overlap_height_ratio,
overlap_width_ratio=overlap_width_ratio,
postprocess_type=postprocess_type,
postprocess_match_metric=postprocess_match_metric,
postprocess_match_threshold=postprocess_match_threshold,
postprocess_class_agnostic=postprocess_class_agnostic,
)
visual_result_2 = sahi.utils.cv.visualize_object_predictions(
image=numpy.array(image),
object_prediction_list=prediction_result_2.object_prediction_list,
)
output = Image.fromarray(visual_result_2["image"])
else:
# standard inference
prediction_result_1 = sahi.predict.get_prediction(
image=image, detection_model=model
)
print(image)
visual_result_1 = sahi.utils.cv.visualize_object_predictions(
image=numpy.array(image),
object_prediction_list=prediction_result_1.object_prediction_list,
)
output = Image.fromarray(visual_result_1["image"])
# sliced inference
return output
inputs = [
gr.Dropdown(choices=["YOLOv5","YOLOv5 + SAHI"],label="Choose Model Type"),
gr.inputs.Image(type="pil", label="Original Image"),
gr.inputs.Number(default=512, label="slice_height"),
gr.inputs.Number(default=512, label="slice_width"),
gr.inputs.Number(default=0.2, label="overlap_height_ratio"),
gr.inputs.Number(default=0.2, label="overlap_width_ratio"),
gr.inputs.Dropdown(
["NMS", "GREEDYNMM"],
type="value",
default="GREEDYNMM",
label="postprocess_type",
),
gr.inputs.Dropdown(
["IOU", "IOS"], type="value", default="IOS", label="postprocess_type"
),
gr.inputs.Number(default=0.5, label="postprocess_match_threshold"),
gr.inputs.Checkbox(default=True, label="postprocess_class_agnostic"),
]
outputs = [
gr.outputs.Image(type="pil", label="Output")
]
title = "Small Object Detection with SAHI + YOLOv5"
description = "SAHI + YOLOv5 demo for small object detection. Upload an image or click an example image to use."
article = "<p style='text-align: center'>SAHI is a lightweight vision library for performing large scale object detection/ instance segmentation.. <a href='https://github.com/obss/sahi'>SAHI Github</a> | <a href='https://medium.com/codable/sahi-a-vision-library-for-performing-sliced-inference-on-large-images-small-objects-c8b086af3b80'>SAHI Blog</a> | <a href='https://github.com/fcakyon/yolov5-pip'>YOLOv5 Github</a> </p>"
examples = [
["apple_tree.jpg", 256, 256, 0.2, 0.2, "GREEDYNMM", "IOS", 0.5, True],
["highway.jpg", 256, 256, 0.2, 0.2, "GREEDYNMM", "IOS", 0.5, True],
["highway2.jpg", 512, 512, 0.2, 0.2, "GREEDYNMM", "IOS", 0.5, True],
["highway3.jpg", 512, 512, 0.2, 0.2, "GREEDYNMM", "IOS", 0.5, True],
]
gr.Interface(
sahi_yolo_inference,
inputs,
outputs,
title=title,
description=description,
article=article,
examples=examples,
theme="huggingface",
).launch(debug=True, enable_queue=True)