TALKtoME: Educational Materials for Speech and Language Acquisition in Autism

This dataset contains image and video samples of action verbs and verb+noun pairs. It is designed to support machine learning tasks related to visual understanding, action recognition, and language grounding.

Dataset Structure

The dataset contains the following main folders:

• action_verbs/images/: image samples organized by action verb categories.
• action_verbs/videos/: video samples organized by action verb categories.
• verb_with_noun/images/: image samples organized by verb+noun categories.
• verb_with_noun/videos/: video samples organized by verb+noun categories.

action_verbs/images

• Total Folders: 50
• Average count: ~10.2 images per verb
• Verbs:

  [
    "bark", "bend over", "bounce", "call", "clean", "color", "comb", "cook", "crawl", "cry", "dance", "dig", "draw", "fall", "fly", "get dressed", "hug", "jump", "kiss", "knock", "laugh", "lie down", "listen", "look", "melt", "paint", "point", "read", "run", "scratch", "sing", "sit", "skate", "ski", "sleep", "slide", "smell", "smile", "spin", "splash", "stir", "stretch", "swim", "swing", "talk", "walk", "wave", "wink", "write", "yawn"
  ]
  

action_verbs/videos

• Total Folders: 50
• Average count: ~4.4 videos per verb
• Verbs:

  [
    "bark", "bend over", "bounce", "call", "clean", "color", "comb", "cook", "crawl", "cry", "dance", "dig", "draw", "fall", "fly", "get dressed", "hug", "jump", "kiss", "knock", "laugh", "lie down", "listen", "look", "melt", "paint", "point", "read", "run", "scratch", "sing", "sit", "skate", "ski", "sleep", "slide", "smell", "smile", "spin", "splash", "stir", "stretch", "swim", "swingle", "talk", "walk", "wave", "wink", "write", "yawn"
  ]
  

verb_with_noun/images

• Total Folders: 113
• Average count: ~4.8 images per verb
• Verbs:

  [
    "bake bread", "bake chicken", "bake pizza", "blow balloon", "blow bubbles", "blow nose", "blow out candles", "bounce the ball", "break board", "break cookie", "brush hair", "brush teeth", "buckle up seat belt", "build blocks", "button shirt", "carry baby", "carry backpack", "carry bag", "carry basket", "carry books", "carry box", "carry bucket", "carry dog", "carry ladder", "catch ball", "clap hands", "climb fence", "climb ladder", "climb rocks", "climb wall", "close box", "comb hair", "cut bread", "cut carrot", "cut cucumber", "cut nails", "cut paper", "dig sand", "drink coffee", "drink juice", "drink milk", "drink tea", "drink water", "drive bus", "drive car", "dry body", "dry hair", "eat apple", "eat banana", "eat cake", "eat chips", "eat corn", "eat ice cream", "kick ball", "make the bed", "pack the bag", "peel apple", "peel banana", "peel egg", "peel orange", "peel potato", "pick up ball", "pick up eggs", "pick up trash", "play game", "play guitar", "play piano", "play soccer", "play trumpet", "play violin", "play with toys", "pour coffee", "pour juice", "pour milk", "pour soup", "pour water", "pull curtain", "pull rope", "pull suitcase", "pull wagon", "push car", "push chair", "push shopping cart", "push stroller", "push wheelbarrow", "read book", "read newspaper", "ride bike", "ride bus", "ride truck", "roll ball", "sail a boat", "scratch arm", "scratch head", "set table", "squeeze lemon", "squeeze orange", "squeeze toothpaste", "sweep floor", "take bath", "throw ball", "tie shoes", "wash apple", "wash car", "wash dishes", "wash hands", "wash tomatoes", "watch tv", "water plants", "wipe floor", "wipe table", "wipe window", "zip jacket"
  ]
  

verb_with_noun/videos

• Total Folders: 136
• Average count: ~3.5 videos per verb
• Verbs:

  [
    "bake bread", "bake chicken", "bake pizza", "blow balloon", "blow bubbles", "blow nose", "blow out candles", "bounce the ball", "break board", "break cookie", "brush hair", "brush teeth", "buckle up seat belt", "build blocks", "button shirt", "carry baby", "carry backpack", "carry bag", "carry basket", "carry books", "carry box", "carry bucket", "carry dog", "carry ladder", "catch ball", "clap hands", "climb fence", "climb ladder", "climb rocks", "climb wall", "close book", "close box", "close door", "close drawer", "close jar", "comb hair", "cut bread", "cut carrot", "cut cucumber", "cut nails", "cut paper", "dig sand", "drink coffee", "drink juice", "drink milk", "drink tea", "drink water", "drive bus", "drive car", "dry body", "dry hair", "eat apple", "eat banana", "eat cake", "eat chips", "eat corn", "eat ice cream", "kick ball", "make the bed", "open book", "open box", "open cabinet", "open door", "open drawer", "open fridge", "open jar", "pack the bag", "peel apple", "peel banana", "peel egg", "peel orange", "peel potato", "pick up ball", "pick up cup", "pick up eggs", "pick up package", "pick up rock", "pick up trash", "play game", "play guitar", "play piano", "play soccer", "play trumpet", "play violin", "play with toys", "pour coffee", "pour juice", "pour milk", "pour soup", "pour water", "pull curtain", "pull rope", "pull suitcase", "pull wagon", "push car", "push chair", "push shopping cart", "push stroller", "push wheelbarrow", "put on glasses", "put on gloves", "put on jacket", "put on shoes", "read book", "read newspaper", "ride bike", "ride bus", "ride truck", "roll ball", "sail a boat", "scratch arm", "scratch head", "set table", "squeeze lemon", "squeeze orange", "squeeze toothpaste", "sweep floor", "take bath", "take off hat", "take off jacket", "take off mask", "take off shoes", "takeoff glasses", "throw ball", "tie shoes", "wash apple", "wash car", "wash dishes", "wash hands", "wash tomatoes", "watch tv", "water plants", "wipe floor", "wipe table", "wipe window", "zip jacket"
  ]
  

Download

You can directly download the fully zipped dataset here:

• Download Entire Dataset (TALKtoME.zip)

Or download specific categories as zip files:

• Download Only Action Verbs (action_verbs.zip)
• Download Only Verb+Noun (verb_with_noun.zip)

Alternatively, you can download the dataset via the Hugging Face CLI:

hf download LSL-datasets/TALKtoME --repo-type dataset

To download a specific folder:

hf download LSL-datasets/TALKtoME --repo-type dataset --include "verb_with_noun/**"

Intended Use

This dataset can be used for training and evaluating models for action recognition, visual-language understanding, and related machine learning tasks.

Limitations

The dataset may not cover all possible variations of each action. Users should be careful when generalizing model performance beyond the dataset distribution.