Split into sister datasets: this one is images only, viewer now auto-detects ImageFolder cleanly

README.md

@@ -4,52 +4,29 @@ tags:
 - gradio
 - examples
 - spaces
-pretty_name: Gradio Space Example Inputs
-configs:
-- config_name: videos
-  data_files:
-  - split: train
-    path: "videos/*.mp4"
-- config_name: images
-  data_files:
-  - split: train
-    path: "images/*.jpg"
+pretty_name: Gradio Space Example Inputs — Images
 ---
 
-# Gradio Space Example Inputs
+# Gradio Space Example Inputs — Images
 
-A small, curated, freely-licensed pool of images and videos used as
-`gr.Examples` for Gradio Spaces that wrap image- and video-generation models.
+A small, curated, freely-licensed pool of images used as `gr.Examples` for
+Gradio Spaces that wrap image-input generation models (image-to-image, edits,
+controls, etc.). Sister dataset for videos:
+[`linoyts/repo-to-space-example-videos`](https://huggingface.co/datasets/linoyts/repo-to-space-example-videos).
 
-When a Space takes media input (image-to-image, image-to-video, video-to-video,
-edits, controls, etc.), the agent building the Space picks 2–3 assets from this
-dataset whose `caption` + `categories` match the model's task, downloads them
-via `hf_hub_download`, runs any model-specific preprocessing (trim / resize to
-expected shapes), and wires them into `gr.Examples`. The Space ships the
-preprocessed copies; this dataset is the source of truth.
+When a Space takes image input, the agent building the Space picks 2–3 images
+whose `caption` + `categories` match the model's task, downloads them via
+`hf_hub_download`, runs any model-specific preprocessing (resize to expected
+shapes), and wires them into `gr.Examples`. The Space ships the preprocessed
+copies; this dataset is the source of truth.
 
-The set is intentionally diverse on subject, framing, motion, and style so a
-caller can pick examples that flatter a specific model without needing to
-source new media each time.
-
-## Layout
-
-Two configs, one per media type, each in its own folder so HF's dataset viewer
-can render real thumbnails. The flat-with-mixed-media layout breaks the viewer
-because `folder_based_builder` only handles one media type per config.
-
-```
-images/
-├── metadata.jsonl     # 19 image records
-└── *.jpg
-videos/
-├── metadata.jsonl     # 15 video records
-└── *.mp4
-```
+The set is intentionally diverse on subject, framing, and style so a caller can
+pick examples that flatter a specific model without needing to source new
+media each time.
 
 ## Schema
 
-Each `metadata.jsonl` — one JSON object per line, one record per asset:
+`metadata.jsonl` — one JSON object per line, one record per asset:
 
 ```json
 {
@@ -66,35 +43,30 @@ Each `metadata.jsonl` — one JSON object per line, one record per asset:
 
 The asset key is `file_name` (with underscore) — required by HF's
 `folder_based_builder` for the dataset viewer; other spellings (`filename`,
-`path`, ...) break it with `SplitsNotFoundError`. The path inside `file_name`
-is relative to the metadata file (just `woman.jpg`, no `images/` prefix).
+`path`, ...) break it with `SplitsNotFoundError`.
 
 | Field | Type | Notes |
 |---|---|---|
-| `file_name` | string | Asset filename, relative to its own subfolder (e.g. `woman.jpg`, not `images/woman.jpg`). |
-| `type` | `"image"` or `"video"` | |
+| `file_name` | string | Filename in this repo root. |
+| `type` | `"image"` | |
 | `width`, `height` | int | Stored dimensions (post the source-shape constraints below). |
-| `duration_s`, `fps` | float | Video only. |
 | `categories` | list of strings | Coarse subject tags for fast filtering. |
-| `caption` | string | Natural-language description of content: subject(s), setting/background, lighting, composition, and (for videos) motion + camera. No use-case judgments — the caller infers fit from the description. |
+| `caption` | string | Natural-language description: subject(s), setting/background, lighting, composition. No use-case judgments — the caller infers fit from the description. |
 | `source`, `license` | string | Provenance / licensing. |
 
-`categories` are for cheap pre-filtering (drop videos for a text-to-image
-model, drop `landscape` for a face-edit model, etc.). `caption` is for the
-finer matching step against the target model's task or example prompts.
+`categories` are for cheap pre-filtering (drop `landscape` for a face-edit
+model, etc.). `caption` is for the finer matching step against the target
+model's task or example prompts.
 
 ## Picking examples for a Space
 
-1. **Pick the config** that matches the model's input modality — `images` or
-   `videos`. Load only that subfolder's `metadata.jsonl`.
-2. **Soft filter by `categories`** — drop tags that are clearly off-task for
-   the target model (e.g. `landscape` for a face-edit model; `dialogue` for a
-   pose-driven motion model).
-3. **Rank by caption fit** — read the surviving captions against the model's
+1. **Soft filter by `categories`** — drop tags that are clearly off-task for
+   the target model (e.g. `landscape` for a face-edit model).
+2. **Rank by caption fit** — read the surviving captions against the model's
    task description, trigger words, or example prompts; pick 2–3 the model
    will plausibly produce a *good* output on, not just any input that doesn't
    crash.
-4. **Diversify the final picks** — different subjects, framings, or lighting
+3. **Diversify the final picks** — different subjects, framings, or lighting
    conditions — so the Examples row teaches users the breadth of what the
    Space handles.
 
@@ -103,15 +75,13 @@ from huggingface_hub import hf_hub_download
 import json
 
 DATASET = "linoyts/repo-to-space-example-inputs"
-# pick the right config based on the model's input modality
-SUBDIR = "images"  # or "videos"
 meta = [json.loads(l) for l in open(
-    hf_hub_download(DATASET, filename=f"{SUBDIR}/metadata.jsonl", repo_type="dataset"))]
+    hf_hub_download(DATASET, filename="metadata.jsonl", repo_type="dataset"))]
 
 # ... filter + rank using `categories` and `caption` ...
 
 chosen = ["woman.jpg", "man_beach.jpg", "girl_with_dog.jpg"]
-paths = [hf_hub_download(DATASET, filename=f"{SUBDIR}/{f}", repo_type="dataset") for f in chosen]
+paths = [hf_hub_download(DATASET, filename=f, repo_type="dataset") for f in chosen]
 # preprocess paths for the target model, then pass to gr.Examples in the Space.
 ```
 
@@ -120,59 +90,9 @@ dataset at Space runtime, and don't set `cache_examples=True` on ZeroGPU.
 
 ## Source-shape constraints
 
-Assets are stored at a sane upper bound:
-
 - **Images:** max 1024 px on the long edge, JPEG quality 90.
-- **Videos:** max 720 px on the short edge, ≤10 s, libx264 CRF 23, no audio.
-
-Callers resize / trim further per model (e.g. LTX wants `8k+1` frames and
-dimensions divisible by 32; Qwen-Image wants multiples of 8/16).
-
-## Contents
-
-**Images (19)**
-
-| Filename | Categories |
-|---|---|
-| anime_1.jpg | illustration, anime, scene |
-| anime_2.jpg | illustration, anime, portrait, character |
-| astronaut.jpg | portrait, person, studio, costume |
-| bunny.jpg | animal, portrait, close-up, nature |
-| cake.jpg | object, food, still-life |
-| dog.jpg | animal, portrait |
-| girl_with_dog.jpg | person, child, animal, outdoor |
-| hand_sketching.jpg | object, close-up, art |
-| horse.jpg | animal, outdoor, landscape |
-| illustration_1.jpg | illustration, scene, children |
-| landscape.jpg | landscape, outdoor |
-| man_beach.jpg | person, portrait, outdoor, beach |
-| mario.jpg | object, still-life, toy |
-| noodles.jpg | object, food, still-life |
-| rubber_ducks.jpg | object, still-life, toy |
-| sign_beach.jpg | object, still-life, outdoor, beach |
-| tent.jpg | object, outdoor, landscape |
-| woman.jpg | portrait, person |
-| woman_sad.jpg | portrait, person, close-up |
-
-**Videos (15)**
-
-| Filename | Categories |
-|---|---|
-| 2_people_running_in_nature.mp4 | motion, person, outdoor, landscape |
-| 2_women_arguing.mp4 | motion, person, dialogue, indoor |
-| brush_mixing_paint.mp4 | motion, object, close-up, art |
-| cat.mp4 | motion, animal, close-up |
-| cat_on_a_tree.mp4 | motion, animal, outdoor |
-| couple_dancing.mp4 | motion, dance, person, indoor |
-| landscape.mp4 | motion, landscape, outdoor |
-| man_dancing.mp4 | motion, dance, person, indoor |
-| man_jumping_rope.mp4 | motion, person, sport, outdoor |
-| man_laughing.mp4 | motion, portrait, person, studio |
-| man_sad.mp4 | motion, portrait, person, studio |
-| podcaster.mp4 | motion, person, indoor |
-| slicing_veggie.mp4 | motion, object, close-up, food |
-| surfing.mp4 | motion, person, sport, outdoor |
-| woman_singing.mp4 | motion, person, music, indoor |
+
+Callers resize further per model (e.g. Qwen-Image wants multiples of 8/16).
 
 ## License
 

videos/2_people_running_in_nature.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:9fce38a331acb21c8caded3a862bcebbb6aa21dd00ef66f2cc76bb177e5f12d2
-size 2690342

videos/2_women_arguing.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:42a9947679545e509c65130af33fbe1eb7459d660a80ba8b566ce5efa5d97386
-size 871673

videos/brush_mixing_paint.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:648228d10e4340f12c840a70cf80e14d1ee150c1311310e1ebf840a3f640b4a8
-size 2476272

videos/cat.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:d22e1330d349cf4da47866d9022709cfc67bc70ad4bdba606cd82fd459c42ce0
-size 778505

videos/cat_on_a_tree.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:e836a1105ea53ad54d6174b117579aec2f88e1b7e3dfe2486e9d80c688181298
-size 986491

videos/couple_dancing.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:45240853f6f0b174a6f80405bafad2b61368e69537f1fa72937ced7c134629fa
-size 990672

videos/landscape.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:434cdcc9bb9cad6ed87437fb3b9ec2de0b3eb48e5c574b8a21e0009fa326285c
-size 4423692

videos/man_dancing.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:11574ac1faa56e0afdfe9bfbb35a26b9f2ea50103f97f93349e6f5375647f0b7
-size 713572

videos/man_jumping_rope.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:53ea745d319bc6268699319e4ca3c680cfe2160f5f002483c4b3fbbe1b9b40b8
-size 1135863

videos/man_laughing.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f3495a689462b346398db9e5767f9aad157634b1b0a4014d706b486a37080b88
-size 1216429

videos/man_sad.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:49471beb5c7c19e981adaf5744a9f221f3990b9c1fb7c17dd1ff0d1014e754d2
-size 885558

videos/metadata.jsonl

@@ -1,15 +0,0 @@
-{"file_name": "2_people_running_in_nature.mp4", "type": "video", "width": 720, "height": 1364, "duration_s": 10.0, "fps": 25.0, "categories": ["motion", "person", "outdoor", "landscape"], "caption": "A young couple — bearded man in a navy puffer vest and white shirt, woman in a striped grey sweater — jog playfully across a moss-covered coastal rock plateau toward the camera; rough sea, a small wind-bent pine and distant rocky shoreline under a flat overcast sky behind them; cool muted light, vertical handheld follow shot with slight camera bob.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "2_women_arguing.mp4", "type": "video", "width": 720, "height": 1280, "duration_s": 5.047, "fps": 23.976, "categories": ["motion", "person", "dialogue", "indoor"], "caption": "Two young women in a bright daylight room with a large frosted window: one with curly hair in a grey turtleneck and jeans stands and points down accusingly at the other, who sits on the windowsill in a brown hoodie with chin on her hand; cool natural window backlight, vertical static medium shot with subtle gestural motion.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "brush_mixing_paint.mp4", "type": "video", "width": 1280, "height": 720, "duration_s": 9.04, "fps": 25.0, "categories": ["motion", "object", "close-up", "art"], "caption": "Extreme close-up of a worn flat-bristle paintbrush swirling a glossy teal-green oil paint mixture on a white palette; rich textured strokes pool and ripple as the brush moves; horizontal shallow-depth shot, blurred art-studio background, warm artificial light, slow circular brush motion.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "cat.mp4", "type": "video", "width": 720, "height": 1280, "duration_s": 4.989, "fps": 59.94, "categories": ["motion", "animal", "close-up"], "caption": "A black-and-white tuxedo cat sits outdoors against an out-of-focus pale wall and grooms its left paw, licking and chewing it with eyes half closed; soft cool overcast light, vertical static close-up shot with minimal camera motion.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "cat_on_a_tree.mp4", "type": "video", "width": 724, "height": 720, "duration_s": 5.0, "fps": 30.0, "categories": ["motion", "animal", "outdoor"], "caption": "A black-and-white spotted cat clings to the trunk of a thin tree on a quiet residential street, ears up and gazing upward; a small tan dog walks past in the background and a parked white car is visible; soft late-afternoon light, near-static vertical handheld shot with slight camera shake.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "couple_dancing.mp4", "type": "video", "width": 720, "height": 1280, "duration_s": 6.006, "fps": 29.97, "categories": ["motion", "dance", "person", "indoor"], "caption": "A young man in a rust-colored t-shirt and a young woman in a yellow top and black leggings perform a contemporary dance lift in a dim rehearsal studio with a wooden floor and dark wall; the woman arches backward with arms extended; warm tungsten lighting, vertical static wide shot.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "landscape.mp4", "type": "video", "width": 1280, "height": 720, "duration_s": 10.01, "fps": 59.94, "categories": ["motion", "landscape", "outdoor"], "caption": "Wide aerial-style daylight view of a green Andean-looking valley with terraced farmland, a small village, dirt tracks, and rolling foothills; tall snow-dappled mountains rise in the distance under a vivid blue sky with scattered white clouds; horizontal slow drone-like sideways pan from a high vantage point, sharp midday light.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "man_dancing.mp4", "type": "video", "width": 720, "height": 1280, "duration_s": 7.007, "fps": 23.976, "categories": ["motion", "dance", "person", "indoor"], "caption": "A bearded man with glasses, tie-dye t-shirt and patchwork harem pants dances expressively alone in an empty industrial warehouse with large grid windows and a dusty concrete floor; soft diffused daylight from the windows behind him, vertical static wide shot, energetic limb motion.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "man_jumping_rope.mp4", "type": "video", "width": 1280, "height": 720, "duration_s": 5.0, "fps": 30.0, "categories": ["motion", "person", "sport", "outdoor"], "caption": "A muscular man in a grey \"COACH\" t-shirt and shorts jumps rope on a city sidewalk in front of a tall arched brick warehouse wall; a stop sign and a Bike Lane sign at left, a utility pole at right; flat midday natural light, horizontal static wide shot, repetitive vertical bouncing motion.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "man_laughing.mp4", "type": "video", "width": 1280, "height": 720, "duration_s": 10.0, "fps": 25.0, "categories": ["motion", "portrait", "person", "studio"], "caption": "Studio close-up portrait of a smiling young Black man in a cream hoodie laughing and looking off-camera, hand near his ear; plain grey backdrop, soft studio key light, horizontal medium static shot with subtle head and shoulder motion.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "man_sad.mp4", "type": "video", "width": 1280, "height": 720, "duration_s": 7.0, "fps": 25.0, "categories": ["motion", "portrait", "person", "studio"], "caption": "Studio close-up portrait of a young man with voluminous brown curly hair, blue eyes, white short-sleeve shirt with floral print over a yellow-striped tee; he covers his nose and mouth with one hand in a downcast expression; plain pale grey background, soft studio light, horizontal medium static shot.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "podcaster.mp4", "type": "video", "width": 1364, "height": 720, "duration_s": 10.0, "fps": 25.0, "categories": ["motion", "person", "indoor"], "caption": "A young woman with long lavender-purple hair, peach over-ear headphones and a beige knit sweater speaks into a large silver condenser microphone on a desk in a cozy home studio; an open MacBook in front of her and a glowing green monstera-leaf neon sign on the wall behind; warm-cool mixed indoor lighting, horizontal medium static shot with light upper-body motion as she talks.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "slicing_veggie.mp4", "type": "video", "width": 1280, "height": 720, "duration_s": 5.96, "fps": 50.0, "categories": ["motion", "object", "close-up", "food"], "caption": "Extreme close-up of a hand using a black-handled chef's knife to slice a green zucchini into thin rounds on a yellow plastic cutting board; bell pepper slices and chopped vegetables blurred in the background of a home kitchen; warm artificial light, horizontal shallow-depth shot, repetitive downward slicing motion.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "surfing.mp4", "type": "video", "width": 720, "height": 1280, "duration_s": 5.68, "fps": 25.0, "categories": ["motion", "person", "sport", "outdoor"], "caption": "A surfer in a black wetsuit rides a small breaking wave on a sunlit ocean with arms raised in celebration as spray fans out around the board; bright blue sky and clean horizon line behind, vertical static long-lens shot from shore, dynamic water motion.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}
-{"file_name": "woman_singing.mp4", "type": "video", "width": 1280, "height": 720, "duration_s": 4.0, "fps": 25.0, "categories": ["motion", "person", "music", "indoor"], "caption": "A young blonde woman with a high ponytail, round tinted pink glasses, layered silver chain necklaces and a white tee sings passionately into a handheld microphone on a dark stage; magenta-and-purple stage lighting wraps the scene, blurred drum kit at far left; horizontal medium close-up static shot with slight gestural motion.", "source": "pexels-or-similar", "license": "free-to-use (CC0-style)"}

videos/podcaster.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:807f28304bb131537247d066e1fa974cc1713f1e2cf8f2f8ebd56d6f897d4d9a
-size 1048990

videos/slicing_veggie.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:dd50308ab3934280e9474b15c6dce541158dd07a0c5f16703210ef338a38b269
-size 871629

videos/surfing.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:3815764aab1dec66f3c48d37274e82918177765c5405bb967041b0dab6df3b0c
-size 2534433

videos/woman_singing.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b94e503cb4d159818e310a1aa28669a668c2eac46fd83e282cd0d8045055f69d
-size 717933