Hugging Face's logo

Spaces:
adalbertojunior
/
image_captioning_portuguese
Runtime error

App Files Community
image_captioning_portuguese
File size: 2,457 Bytes 
f8bd957
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d08e99b
f8bd957
 
 
 
 
 
 
e50de60
edf9b2b
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
 
import streamlit as st
from transformers import VisionEncoderDecoderModel, AutoFeatureExtractor, AutoTokenizer
import requests
from PIL import Image
import torch


CHECKPOINT = "adalbertojunior/image_captioning_portuguese"

@st.cache
def get_model():
    model = VisionEncoderDecoderModel.from_pretrained(CHECKPOINT)
    return model


feature_extractor = AutoFeatureExtractor.from_pretrained(CHECKPOINT)
tokenizer = AutoTokenizer.from_pretrained(CHECKPOINT)

st.title("Image Captioning with ViT & GPT2 🇧🇷")

st.sidebar.markdown("## Generation parameters")
max_length = st.sidebar.number_input("Max length", value=20, min_value=1)
no_repeat_ngram_size = st.sidebar.number_input("no repeat ngrams size", value=2, min_value=1)
num_return_sequences = st.sidebar.number_input("Generated sequences", value=3, min_value=1)

gen_mode = st.sidebar.selectbox("Generation mode", ["beam search", "sampling"])
if gen_mode == "beam search": 
    num_beams = st.sidebar.number_input("Beam size", value=5, min_value=1)
    early_stopping = st.sidebar.checkbox("Early stopping", value=True)
    gen_params = {
        "num_beams": num_beams,
        "early_stopping": early_stopping
    }
elif gen_mode == "sampling":
    do_sample = True
    top_k = st.sidebar.number_input("top_k", value=30, min_value=0)
    top_p = st.sidebar.number_input("top_p", value=0, min_value=0)  
    temperature = st.sidebar.number_input("temperature", value=0.7, min_value=0.0)
    gen_params = {
        "do_sample": do_sample,
        "top_k": top_k,
        "top_p": top_p,
        "temperature": temperature
    }

def generate_caption(url):
    image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
    inputs = feature_extractor(image, return_tensors="pt")
    model = get_model()
    model.eval()
    generated_ids = model.generate(
        inputs["pixel_values"],
        max_length=20,
        no_repeat_ngram_size=2,
        num_return_sequences=3,
        **gen_params
    )
    captions = tokenizer.batch_decode(
        generated_ids,
        skip_special_tokens=True,
    )
    return captions[0]


url = st.text_input(
    "Insert your URL", "https://static.cdn.pleno.news/2017/09/avi%C3%A3o-e1572374124339.jpg"
)

st.image(url)

if st.button("Run captioning"):
    with st.spinner("Processing image..."):
        caption = generate_caption(url)
        st.text(caption)
st.text("Research supported with Cloud TPUs from Google's TPU Research Cloud (TRC)")