Hugging Face's logo

Spaces:
flax-community
/
spanish-image-captioning
Runtime error

App Files Community
spanish-image-captioning
File size: 4,951 Bytes 
3a2e60d
 
 
 
 
 
 
 
84c69e6
3a2e60d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
053cee9
3a2e60d
 
 
 
 
 
 
 
 
 
 
 
 
 
053cee9
 
 
3a2e60d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
 
from io import BytesIO
import streamlit as st
import pandas as pd
import json
import os
import numpy as np
from streamlit.elements import markdown
from PIL import Image
from model.flax_clip_vision_marian.modeling_clip_vision_marian import (
    FlaxCLIPVisionMarianMT,
)
from transformers import MarianTokenizer
from utils import (
    get_transformed_image,
)
import matplotlib.pyplot as plt
from mtranslate import translate


from session import _get_state

state = _get_state()


@st.cache
def load_model(ckpt):
    return FlaxCLIPVisionMarianMT.from_pretrained(ckpt)


tokenizer = MarianTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-es")

@st.cache(persist=True)
def generate_sequence(pixel_values, num_beams, temperature, top_p):
    output_ids = model.generate(input_ids=pixel_values, max_length=64, num_beams=num_beams, temperature=temperature, top_p = top_p)
    print(output_ids)
    output_sequence = tokenizer.batch_decode(output_ids[0], skip_special_tokens=True, max_length=64)
    return output_sequence

def read_markdown(path, parent="./sections/"):
    with open(os.path.join(parent, path)) as f:
        return f.read()


checkpoints = ["./ckpt/ckpt-17499"]  # TODO: Maybe add more checkpoints?
dummy_data = pd.read_csv("references.tsv", sep="\t")

st.set_page_config(
    page_title="Multilingual Image Captioning",
    layout="wide",
    initial_sidebar_state="collapsed",
    page_icon="./misc/mic-logo.png",
)

st.title("Multilingual Image Captioning")
st.write(
    "[Bhavitvya Malik](https://huggingface.co/bhavitvyamalik), [Gunjan Chhablani](https://huggingface.co/gchhablani)"
)

st.sidebar.title("Generation Parameters")
num_beams = st.sidebar.number_input("Number of Beams", min_value=2, max_value=10, value=4, step=1, help="Number of beams to be used in beam search.")
temperature = st.sidebar.select_slider("Temperature", options = list(np.arange(0.0,1.1, step=0.1)), value=1.0, help ="The value used to module the next token probabilities.", format_func=lambda x: f"{x:.2f}")
top_p = st.sidebar.select_slider("Top-P", options = list(np.arange(0.0,1.1, step=0.1)),value=1.0, help="Nucleus Sampling : If set to float < 1, only the most probable tokens with probabilities that add up to :obj:`top_p` or higher are kept for generation.", format_func=lambda x: f"{x:.2f}")


image_col, intro_col = st.beta_columns([3, 8])
# image_col.image("./misc/sic-logo.png", use_column_width="always")
intro_col.write(read_markdown("intro.md"))

with st.beta_expander("Usage"):
    st.markdown(read_markdown("usage.md"))

with st.beta_expander("Article"):
    st.write(read_markdown("abstract.md"))
    st.write(read_markdown("caveats.md"))
    # st.write("# Methodology")
    # st.image(
    #     "./misc/Multilingual-IC.png", caption="Seq2Seq model for Image-text Captioning."
    # )
    st.markdown(read_markdown("pretraining.md"))
    st.write(read_markdown("challenges.md"))
    st.write(read_markdown("social_impact.md"))
    st.write(read_markdown("references.md"))
    # st.write(read_markdown("checkpoints.md"))
    st.write(read_markdown("acknowledgements.md"))


first_index = 20
# Init Session State
if state.image_file is None:
    state.image_file = dummy_data.loc[first_index, "image_file"]
    state.caption = dummy_data.loc[first_index, "caption"].strip("- ")

    image_path = os.path.join("images", state.image_file)
    image = plt.imread(image_path)
    state.image = image

# col1, col2 = st.beta_columns([6, 4])

if st.button("Get a random example", help="Get a random example from one of the seeded examples."):
    sample = dummy_data.sample(1).reset_index()
    state.image_file = sample.loc[0, "image_file"]
    state.caption = sample.loc[0, "caption"].strip("- ")

    image_path = os.path.join("images", state.image_file)
    image = plt.imread(image_path)
    state.image = image

# col2.write("OR")

# uploaded_file = col2.file_uploader("Upload your image", type=["png", "jpg", "jpeg"])
# if uploaded_file is not None:
#     state.image_file = os.path.join("images", uploaded_file.name)
#     state.image = np.array(Image.open(uploaded_file))

transformed_image = get_transformed_image(state.image)

new_col1, new_col2 = st.beta_columns([5,5])
# Display Image
new_col1.image(state.image, use_column_width="always")


# Display Reference Caption
new_col2.write("**Reference Caption**: " + state.caption)
new_col2.markdown(
    f"""**English Translation**: {translate(state.caption, 'en')}"""
)

with st.spinner("Loading model..."):
    model = load_model(checkpoints[0])
sequence = ['']
if new_col2.button("Generate Caption", help="Generate a caption in the specified language."):
    with st.spinner("Generating Sequence..."):
        sequence = generate_sequence(transformed_image, num_beams, temperature, top_p)
# print(sequence)

if sequence!=['']:
    st.write(
        "**Generated Caption**: "+sequence[0]
    )

    st.write(
        "**English Translation**: "+  translate(sequence[0])
    )