WIP

app.py

@@ -18,6 +18,7 @@ load_dotenv()
 it_dataset = (
     load_dataset("imagenet-1k", split="train", streaming=True, trust_remote_code=True)
     .shuffle(42)
+    .skip(0)
     .take(1000)
 )
 
@@ -58,14 +59,13 @@ with open("imagenet_categories_data.json") as f:
         filter(lambda x: x["name"] == "artifact, artefact", broad_categories)
     )
     artifact_categories = list(find_categories(artifact_category))
-    logger.info(f"Artifact categories: {artifact_categories}")
+    # logger.info(f"Artifact categories: {artifact_categories}")
 
 
 def filter_imgs_by_label(x):
     """
     Filter out the images that have label -1
     """
-    logger.info(f'label: {x["label"]} (present: {x["label"] in artifact_categories})')
     return x["label"] in artifact_categories
 
 
@@ -98,12 +98,35 @@ def get_user_prompt():
 
 
 hf_writer = gr.HuggingFaceDatasetSaver(
-    hf_token=os.environ["HF_TOKEN"], dataset_name="acmc/maker-faire-bot", private=True
+    hf_token=os.environ["HF_TOKEN"],
+    dataset_name="acmc/maker-faire-bot",
+    private=True,
 )
 csv_writer = gr.CSVLogger()
 
 theme = gr.themes.Default(primary_hue="cyan", secondary_hue="fuchsia")
 
+translation_table = {
+    "Maker Faire Bot": "Maker Faire Bot",
+    "**Think about these objects...**": "**Tænk på disse objekter...**",
+    "We want to build a Maker Faire Bot that can generate creative ideas. Help us by providing ideas on what you'd build with the following three objects!": "Vi vil bygge en Maker Faire Bot, der kan generere kreative ideer. Hjælp os ved at give ideer til, hvad du ville bygge med de følgende tre objekter!",
+    "Change": "Skift",
+    "What would you build with these 3 things?": "Hvad ville du bygge med disse 3 ting?",
+    "For example, if you have a roll of string, a camera, and a loudspeaker, you could build an electronic guitar. If you can write in Danish, that's great!": "For eksempel, hvis du har en rulle snor, et kamera og en højttaler, kunne du bygge en elektronisk guitar. Hvis du kan skrive på dansk, er det fantastisk!",
+    "It doesn't need to be a very long explanation, just a few sentences to help the bot understand your idea.": "Det behøver ikke være en meget lang forklaring, bare et par sætninger for at hjælpe robotten med at forstå din idé.",
+    "Submit": "Indsend",
+    "New Prompt": "Ny opgave",
+    "How would you build it?": "Hvordan ville du bygge det?",
+    "This is an experimental project. Your data is anonymous and will be used to train an AI model. By using this tool, you agree to our policy.": "Dette er et eksperimentelt projekt. Dine data er anonyme og vil blive brugt til at træne en AI-model. Ved at bruge dette værktøj accepterer du vores politik.",
+    "(example): An digital electronic guitar": "(eksempel): En digital elektronisk guitar",
+    """I would use the roll of string to create the strings of the guitar, and the camera to analyze the hand movements. Then, I would use an AI model to predict the chords and play the sound through the loudspeaker.""": """Jeg ville bruge snoren til at skabe guitarens strenge, og kameraet til at analysere håndbevægelserne. Derefter ville jeg bruge en AI-model til at forudsige akkorderne og afspille lyden gennem højttaleren.""",
+}
+
+
+def get_bilingual_string(key):
+    return f"{translation_table[key]}  //  {key}"
+
+
 with gr.Blocks(theme=theme) as demo:
     with gr.Row() as header:
         gr.Image(
@@ -116,73 +139,109 @@ with gr.Blocks(theme=theme) as demo:
             scale=0.2,
         )
         gr.Markdown(
-            """
-            # Maker Faire Bot
-            """,
+            get_bilingual_string("Maker Faire Bot"),
             visible=False,
         )
 
     user_prompt = gr.State(get_user_prompt())
 
-    gr.Markdown("""# Think about these objects...""")
+    gr.Markdown(get_bilingual_string("**Think about these objects...**"))
     gr.Markdown(
-        """We want to teach the Maker Faire Bot some creativity. Help us get ideas on what you'd build!"""
+        get_bilingual_string(
+            "We want to build a Maker Faire Bot that can generate creative ideas. Help us by providing ideas on what you'd build with the following three objects!"
+        )
     )
     image_components = []
     with gr.Row(variant="panel") as row:
-        for i in range(len(user_prompt.value["images"])):
-            with gr.Column(variant="default") as col:
-                img = gr.Image(
-                    user_prompt.value["images"][i],
-                    label=user_prompt.value["human_labels"][i],
-                    interactive=False,
-                    show_download_button=False,
-                    show_share_button=False,
-                )
-                image_components.append(img)
-                btn = gr.Button("Change", variant="secondary")
-
-                def change_image(user_prompt):
-                    logger.info(f"Current user prompt: {user_prompt}")
-                    data = dataset[random.randint(0, len(dataset) - 1)]
-                    user_prompt = user_prompt.copy()
-                    user_prompt["images"][i] = data["image"]
-                    user_prompt["machine_labels"][i] = data["label"]
-                    user_prompt["human_labels"][i] = dataset.features["label"].int2str(
-                        data["label"]
-                    )
-                    return (
-                        user_prompt,
-                        user_prompt["images"][i],
-                        gr.update(
-                            label=user_prompt["human_labels"][i],
-                        ),
-                    )
-
-                btn.click(
-                    change_image,
-                    inputs=[user_prompt],
-                    outputs=[user_prompt, img, img],
-                    preprocess=True,
-                    postprocess=True,
-                )
+        def change_image(this_i, user_prompt):
+            logger.info(
+                f"Current user prompt: {user_prompt}, current image index: {this_i}"
+            )
+            data = dataset[random.randint(0, len(dataset) - 1)]
+            new_user_prompt = user_prompt.copy()
+            new_user_prompt["images"][this_i] = data["image"]
+            new_user_prompt["machine_labels"][this_i] = data["label"]
+            new_user_prompt["human_labels"][this_i] = dataset.features[
+                "label"
+            ].int2str(data["label"])
+            logger.info(f"New user prompt: {new_user_prompt}")
+            return (
+                new_user_prompt,
+                new_user_prompt["images"][this_i],
+                gr.update(
+                    label=new_user_prompt["human_labels"][this_i],
+                ),
+            )
+
+        with gr.Column(variant="default") as col:
+            img = gr.Image(
+                user_prompt.value["images"][0],
+                label=user_prompt.value["human_labels"][0],
+                interactive=False,
+                show_download_button=False,
+                show_share_button=False,
+            )
+            image_components.append(img)
+            btn = gr.Button(get_bilingual_string("Change"), variant="secondary")
+
+            btn.click(
+                lambda *args: change_image(0, *args),
+                inputs=[user_prompt],
+                outputs=[user_prompt, img, img],
+                preprocess=True,
+                postprocess=True,
+            )
+        with gr.Column(variant="default") as col:
+            img = gr.Image(
+                user_prompt.value["images"][1],
+                label=user_prompt.value["human_labels"][1],
+                interactive=False,
+                show_download_button=False,
+                show_share_button=False,
+            )
+            image_components.append(img)
+            btn = gr.Button(get_bilingual_string("Change"), variant="secondary")
+
+            btn.click(
+                lambda *args: change_image(1, *args),
+                inputs=[user_prompt],
+                outputs=[user_prompt, img, img],
+                preprocess=True,
+                postprocess=True,
+            )
+        with gr.Column(variant="default") as col:
+            img = gr.Image(
+                user_prompt.value["images"][2],
+                label=user_prompt.value["human_labels"][2],
+                interactive=False,
+                show_download_button=False,
+                show_share_button=False,
+            )
+            image_components.append(img)
+            btn = gr.Button(get_bilingual_string("Change"), variant="secondary")
+
+            btn.click(
+                lambda *args: change_image(2, *args),
+                inputs=[user_prompt],
+                outputs=[user_prompt, img, img],
+                preprocess=True,
+                postprocess=True,
+            )
 
     user_answer_object = gr.Textbox(
         autofocus=True,
-        placeholder="(example): An digital electronic guitar",
-        label="What would you build?",
+        placeholder=get_bilingual_string("(example): An digital electronic guitar"),
+        label=get_bilingual_string("What would you build with these 3 things?"),
+        info=get_bilingual_string("For example, if you have a roll of string, a camera, and a loudspeaker, you could build an electronic guitar. If you can write in Danish, that's great!")
     )
     user_answer_explanation = gr.TextArea(
         autofocus=True,
-        label="How would you build it?",
+        label=get_bilingual_string("How would you build it?"),
         # The example uses a roll of string, a camera, and a loudspeaker to build an electronic guitar.
-        placeholder="""To build an electronic guitar, I would:
-1. Use the roll of string to create the strings of the guitar.
-2. Use the camera to capture a live video of the hand movements. That way, I can use an AI model to predict the chords.
-3. Using a computer vision model, identify where the fingers are placed on the strings.
-4. Calculate the sounds that the loudspeaker should produce based on the finger placements.
-5. Play the sound through the loudspeaker.
-""",
+        placeholder=get_bilingual_string(
+            """I would use the roll of string to create the strings of the guitar, and the camera to analyze the hand movements. Then, I would use an AI model to predict the chords and play the sound through the loudspeaker."""
+        ),
+        info=get_bilingual_string("It doesn't need to be a very long explanation, just a few sentences to help the bot understand your idea.")
     )
 
     csv_writer.setup(
@@ -194,20 +253,20 @@ with gr.Blocks(theme=theme) as demo:
         flagging_dir="user_data_hf",
     )
 
-    submit_btn = gr.Button("Submit", variant="primary")
+    submit_btn = gr.Button(get_bilingual_string("Submit"), variant="primary")
 
     def log_results(prompt, object, explanation):
         logger.info(f"logging - Prompt: {prompt}")
-        csv_writer.flag(
-            [
-                {
-                    "machine_labels": prompt["machine_labels"],
-                    "human_labels": prompt["human_labels"],
-                },
-                object,
-                explanation,
-            ]
-        )
+        # csv_writer.flag(
+        #    [
+        #        {
+        #            "machine_labels": prompt["machine_labels"],
+        #            "human_labels": prompt["human_labels"],
+        #        },
+        #        object,
+        #        explanation,
+        #    ]
+        # )
         hf_writer.flag(
             [
                 {
@@ -218,25 +277,36 @@ with gr.Blocks(theme=theme) as demo:
                 explanation,
             ]
         )
+        return ["", ""]  # Clear the textboxes
 
     submit_btn.click(
         log_results,
         inputs=[user_prompt, user_answer_object, user_answer_explanation],
+        outputs=[user_answer_object, user_answer_explanation],
         preprocess=True,
     )
 
-    new_prompt_btn = gr.Button("New Prompt", variant="secondary")
-    new_prompt_btn.click(
-        get_user_prompt,
-        outputs=[user_prompt],
-        # preprocess=True,
-    )
+    # def renew_prompt(image_components):
+    #    new_prompt = get_user_prompt()
+    #    for i in range(len(new_prompt["images"])):
+    #        image_components[i].update(
+    #            url=new_prompt["images"][i],
+    #            label=new_prompt["human_labels"][i],
+    #        )
+    #    return new_prompt
+
+    # new_prompt_btn = gr.Button(get_bilingual_string("New Prompt"), variant="secondary")
+    # new_prompt_btn.click(
+    #    renew_prompt,
+    #    inputs=image_components,
+    #    outputs=[user_prompt],
+    #    # preprocess=True,
+    # )
 
     gr.Markdown(
-        """
-        This is an experimental project. Your data is anonymous and will be used to train an AI model. By using this tool, you agree to our [policy](https://makerfaire.com/privacy).
-        """
+        get_bilingual_string(
+            "This is an experimental project. Your data is anonymous and will be used to train an AI model. By using this tool, you agree to our policy."
+        )
     )
-
 if __name__ == "__main__":
     demo.launch()

requirements.txt

@@ -1,2 +1,3 @@
 datasets==2.19.0
 python-dotenv==1.0.1
+gradio==4.28.0