app.py

# from PIL import Image
from transformers import DetrFeatureExtractor
from transformers import DetrForObjectDetection
import torch
# import numpy as np

def object_count(picture):

    feature_extractor = DetrFeatureExtractor.from_pretrained("facebook/detr-resnet-101-dc5")
    encoding = feature_extractor(picture, return_tensors="pt")
    model = DetrForObjectDetection.from_pretrained("facebook/detr-resnet-101-dc5")
    outputs = model(**encoding)
    # keep only predictions of queries with 0.9+ confidence (excluding no-object class)
    probas = outputs.logits.softmax(-1)[0, :, :-1]
    keep = probas.max(-1).values > 0.7
    count = 0
    for i in keep:
        if i:
            count=count+1

    return "About " + str(count) +" common objects were detected"

# object_count("toothbrush.jpg")
import gradio as gr

interface = gr.Interface(object_count, gr.inputs.Image(shape=(640, 480)), "text", title="Common Object Counter",examples=["chairs.jpg", "empty.jpg", "bottles.jpg"], description="This App counts the common objects detected in an Image",
             allow_flagging="never").launch()