import streamlit as st from ispice import Spice # Sample example captions sample_examples = { "ID - Example 1": { "generated": "P1 sits on the couch. P1 puts her hands on the couch and sits up. P2 sits beside her.", "reference": "P1 straightens her posture. P1 sits up. P2 sits beside her on the couch." } } # Function to compute score def preprocess_captions(generated_captions, reference_captions): hypotheses = {'image'+str(i): [generated_captions[i]] for i in range(len(generated_captions))} references = {'image'+str(i): [reference_captions[i]] for i in range(len(reference_captions))} return hypotheses, references # Streamlit app def main(): st.title("iSPICE Metric Evaluation") # Dropdown for comparison option mode = st.selectbox("Mode:", ["ID", "Name"]) spice_scorer = Spice(mode=mode) # Note on how to use the app id_label = r'''If you choose the Mode as ID, supported Person IDs are from P1, P2, P3 .. upto P11.''' name_label = r'''If you choose the Mode as Name, supported Person Names are - Ray, Sam, Casey, Riley, Morgan, Alex, Quinn, Cameron, Avery, Charlie, Jamie and Mike.''' st.caption(id_label) st.caption(name_label) # Sample examples dropdown st.sidebar.subheader("Choose Sample Example:") example_choice = st.sidebar.selectbox("Select Example", list(sample_examples.keys())) # Description st.write("You can either input single caption or multiple captions separated by new line.") generated_caption = sample_examples[example_choice]["generated"] reference_caption = sample_examples[example_choice]["reference"] generated_caption_input = st.text_area("Generated Caption:", value=generated_caption) reference_caption_input = st.text_area("Reference Caption:", value=reference_caption) st.caption("We have a default caption here you can try out. You can modify it as required.") # Compute score button if st.button("Compute Score"): generated_captions = generated_caption_input.split("\n") reference_captions = reference_caption_input.split("\n") print(generated_captions, len(generated_captions)) print(reference_captions, len(reference_captions)) hypotheses, references = preprocess_captions(generated_captions, reference_captions) if generated_caption.strip() == "" or reference_caption.strip() == "": st.error("Please provide both generated and reference captions.") else: average_spice_score, spice_scores, average_ispice_score, ispice_scores = spice_scorer.compute_score(references, hypotheses) st.subheader("Scores :") st.write("Average SPICE Score:", average_spice_score) st.write("Average iSPICE Score:", average_ispice_score) st.write("SPICE Scores:", spice_scores) st.write("iSPICE Scores:", ispice_scores) if __name__ == "__main__": main()