Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -45,6 +45,7 @@ clip_processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
|
|
45 |
|
46 |
# Load the Unsplash dataset
|
47 |
dataset = load_dataset("jamescalam/unsplash-25k-photos", split="train") # all 25K images are in train split
|
|
|
48 |
|
49 |
# Load gpt and modifed weights for captions
|
50 |
gpt = GPT2LMHeadModel.from_pretrained('gpt2')
|
@@ -229,7 +230,7 @@ with gr.Blocks() as demo:
|
|
229 |
1. Enter list of labels separated by commas (or select one of the examples below)
|
230 |
2. Click **Get Random Image** to grab a random image from dataset
|
231 |
3. Click **Classify Image** to analyze current image against the labels (including after changing labels)
|
232 |
-
|
233 |
gr.Markdown(instructions)
|
234 |
with gr.Row(variant="compact"):
|
235 |
label_text = gr.Textbox(show_label=False, placeholder="Enter classification labels").style(container=False)
|
@@ -253,15 +254,17 @@ with gr.Blocks() as demo:
|
|
253 |
get_btn.click(fn=rand_image, outputs=im)
|
254 |
#im.change(predict, inputs=[im, labels], outputs=cf)
|
255 |
class_btn.click(predict, inputs=[im, labels], outputs=cf)
|
|
|
|
|
256 |
|
257 |
with gr.Tab("Captioning"):
|
258 |
instructions = """## Instructions:
|
259 |
1. Click **Get Random Image** to grab a random image from dataset
|
260 |
1. Click **Create Caption** to generate a caption for the image
|
261 |
1. Different models can be selected:
|
262 |
-
|
263 |
-
|
264 |
-
|
265 |
gr.Markdown(instructions)
|
266 |
with gr.Row():
|
267 |
with gr.Column(variant="panel"):
|
@@ -274,12 +277,13 @@ with gr.Blocks() as demo:
|
|
274 |
get_btn_cap.click(fn=rand_image, outputs=im_cap)
|
275 |
#im_cap.change(generate_text, inputs=im_cap, outputs=caption)
|
276 |
caption_btn.click(get_caption, inputs=[im_cap, model_name], outputs=caption)
|
|
|
277 |
|
278 |
with gr.Tab("Search"):
|
279 |
instructions = """## Instructions:
|
280 |
1. Enter a search query (or select one of the examples below)
|
281 |
2. Click **Find Images** to find images that match the query (top 5 are shown in order from left to right)
|
282 |
-
3.
|
283 |
gr.Markdown(instructions)
|
284 |
with gr.Column(variant="panel"):
|
285 |
desc = gr.Textbox(show_label=False, placeholder="Enter description").style(container=False)
|
@@ -292,5 +296,6 @@ with gr.Blocks() as demo:
|
|
292 |
search_btn = gr.Button("Find Images").style(full_width=False)
|
293 |
gallery = gr.Gallery(show_label=False).style(grid=(2,2,3,5))
|
294 |
search_btn.click(search,inputs=desc, outputs=gallery, postprocess=False)
|
|
|
295 |
|
296 |
demo.launch()
|
|
|
45 |
|
46 |
# Load the Unsplash dataset
|
47 |
dataset = load_dataset("jamescalam/unsplash-25k-photos", split="train") # all 25K images are in train split
|
48 |
+
dataset_size = len(dataset)
|
49 |
|
50 |
# Load gpt and modifed weights for captions
|
51 |
gpt = GPT2LMHeadModel.from_pretrained('gpt2')
|
|
|
230 |
1. Enter list of labels separated by commas (or select one of the examples below)
|
231 |
2. Click **Get Random Image** to grab a random image from dataset
|
232 |
3. Click **Classify Image** to analyze current image against the labels (including after changing labels)
|
233 |
+
"""
|
234 |
gr.Markdown(instructions)
|
235 |
with gr.Row(variant="compact"):
|
236 |
label_text = gr.Textbox(show_label=False, placeholder="Enter classification labels").style(container=False)
|
|
|
254 |
get_btn.click(fn=rand_image, outputs=im)
|
255 |
#im.change(predict, inputs=[im, labels], outputs=cf)
|
256 |
class_btn.click(predict, inputs=[im, labels], outputs=cf)
|
257 |
+
gr.HTML(f"Dataset: <a href='https://github.com/unsplash/datasets' target='_blank'>Unsplash Lite</a>; Number of Images: {dataset_size}")
|
258 |
+
|
259 |
|
260 |
with gr.Tab("Captioning"):
|
261 |
instructions = """## Instructions:
|
262 |
1. Click **Get Random Image** to grab a random image from dataset
|
263 |
1. Click **Create Caption** to generate a caption for the image
|
264 |
1. Different models can be selected:
|
265 |
+
* **COCO** generally produces more straight-forward captions, but it is a smaller dataset and therefore struggles to recognize certain objects
|
266 |
+
* **Conceptual Captions** is a much larger dataset but sometimes produces results that resemble social media posts
|
267 |
+
"""
|
268 |
gr.Markdown(instructions)
|
269 |
with gr.Row():
|
270 |
with gr.Column(variant="panel"):
|
|
|
277 |
get_btn_cap.click(fn=rand_image, outputs=im_cap)
|
278 |
#im_cap.change(generate_text, inputs=im_cap, outputs=caption)
|
279 |
caption_btn.click(get_caption, inputs=[im_cap, model_name], outputs=caption)
|
280 |
+
gr.HTML(f"Dataset: <a href='https://github.com/unsplash/datasets' target='_blank'>Unsplash Lite</a>; Number of Images: {dataset_size}")
|
281 |
|
282 |
with gr.Tab("Search"):
|
283 |
instructions = """## Instructions:
|
284 |
1. Enter a search query (or select one of the examples below)
|
285 |
2. Click **Find Images** to find images that match the query (top 5 are shown in order from left to right)
|
286 |
+
3. Keep in mind that the dataset contains mostly nature-focused images"""
|
287 |
gr.Markdown(instructions)
|
288 |
with gr.Column(variant="panel"):
|
289 |
desc = gr.Textbox(show_label=False, placeholder="Enter description").style(container=False)
|
|
|
296 |
search_btn = gr.Button("Find Images").style(full_width=False)
|
297 |
gallery = gr.Gallery(show_label=False).style(grid=(2,2,3,5))
|
298 |
search_btn.click(search,inputs=desc, outputs=gallery, postprocess=False)
|
299 |
+
gr.HTML(f"Dataset: <a href='https://github.com/unsplash/datasets' target='_blank'>Unsplash Lite</a>; Number of Images: {dataset_size}")
|
300 |
|
301 |
demo.launch()
|