Dataset Viewer
Auto-converted to Parquet Duplicate
video_id
large_stringlengths
25
44
sentence
large_stringlengths
3
977
features
unknown
shape
listlengths
3
3
--7E2sU6zP4_10-5-rgb_front
"And I call them decorative elements because basically all they're meant to do is to enrich and colo(...TRUNCATED)
"0vUhvl+5Ub/sRpW/QKPqPrMoGjwhh0y+yxUYP2GUsj/XCLu/J7Qnv5Cypj+cG5G/RaTQvepZxz4/GCvAno9PPZdItz2QS0LAFVs(...TRUNCATED)
[ 365, 96, 3 ]
--7E2sU6zP4_11-5-rgb_front
"So they don't really have much of a symbolic meaning other than maybe life is richer, life is beaut(...TRUNCATED)
"H66fvf2BSL8RIQTAfpj8Pph9Cj0tWpc9lz1TP95SqD//DBo/QuZZvx/DmT93idm/KyKyvlrtwD5apYHAjRlcP3A+M0BNgyo+Ci7(...TRUNCATED)
[ 565, 96, 3 ]
--7E2sU6zP4_12-5-rgb_front
"Now this is very, this is actually an insert of a kind of an envelope for stationary, and this is a(...TRUNCATED)
"P2P9vTkHaL/jcsO/ZIncPqu/yjr6/oG+TnQ6Pypdqj++7s6/SA5Mv7MFhT89pKO/TxTVvrUQUD7V8FzAeD4EvUQiBT+913TA8bk(...TRUNCATED)
[ 349, 96, 3 ]
--7E2sU6zP4_13-5-rgb_front
This is all the you know, take off on the idea of the acanthus leaf.
"tZgRvTICfb/L6fq/Rh72PoNOXTlF4Qy+OiFLPwvHrj+/x3W//BFFvxOQoT/dj6G9Kt/FvlWEA0B9Ns+/KLlCPvbwZj9RsVHAw3p(...TRUNCATED)
[ 144, 96, 3 ]
--7E2sU6zP4_5-5-rgb_front
It's almost has a feathery like posture to it.
"u9sYvsWuf78HUre/fIj3PsLLVzzl4wG+ddRGPy+gtT8v+8S/WQI7v3omrz/blqS/7rL8voKtmz5Tjl/AtlwFP0xWUD/1TH/AgYk(...TRUNCATED)
[ 212, 96, 3 ]
--7E2sU6zP4_6-5-rgb_front
"And so, it's used in architecture as a decorative element in architecture on columns and so on, and(...TRUNCATED)
"JTdDvbZLdb+CRNK/C6z5PvVaAr1jrti9Nw5JPzqUtz8yT/K/8EVNv8o8pz8LXuC/Kwzavm0UUT6jgn7AGNcRPUBMhT43F4/AnBS(...TRUNCATED)
[ 402, 96, 3 ]
--7E2sU6zP4_7-5-rgb_front
"And so what's happened with the idea of acanthus leaf, is that it has is taken on all these differe(...TRUNCATED)
"Y1X8vQc0Ur+S0re/LETUPlmOebxt5I6+425APy2mmj+KjMq/BwFKv2bVgT9Tbpe/vY/avmipFj4eYEzAlDXLvdeBCz/bfF/AddY(...TRUNCATED)
[ 289, 96, 3 ]
--7E2sU6zP4_8-5-rgb_front
"And it has been wildly colored so you can look at some an acanthus leaf and you can start with the (...TRUNCATED)
"xhpwvVdPNL8yS6a/4pfCPrmBvjuOUKa+wPBZPy9gEz9Br7S/8qVVv6Kv+z6a6jS/RvpDv00X5r4wqBjAMVBNPqdsDj9eS0DAt8T(...TRUNCATED)
[ 364, 96, 3 ]
--7E2sU6zP4_9-5-rgb_front
"Here, actually, I have some samples of traced acanthus leaves from different illuminated manuscript(...TRUNCATED)
"NhsXvjTHeL9O5cC/GKnrPkV1rbyH0ka+i2xHP4LboD+CXta/9GI6v45/nj+iOcK/dO90vnMdAj+K7mrA8C3Xuz0IPz81MG3AdqU(...TRUNCATED)
[ 239, 96, 3 ]
--8pSDeC-fg_0-5-rgb_front
Hi.
"O3ICvrr8gr/CbOO/sGz0Ps2HDz3Z7RO+5GAuP2vmnz/tnCO+KpU+v+2gnD/8OBg+nabsvgXVEkCZbz+/5UQIPm9MJECrHai/z3E(...TRUNCATED)
[ 15, 96, 3 ]
End of preview. Expand in Data Studio

A specially selected set of 96 mediapipe coordinates from Mediapipe Holistic.

image

import numpy as np
from datasets import load_dataset

dataset_name = "bdanko/how2sign-landmarks-curated-normalized"
print(f"Loading dataset {dataset_name}...")
dataset = load_dataset(dataset_name, split="test", streaming=True)

# Get a sample
sample = next(iter(dataset))
video_id = sample['video_id']
sentence = sample['sentence']
features_raw = sample['features']
shape = sample['shape']

# Decode features
landmarks = np.frombuffer(features_raw, dtype=np.float32).reshape(shape)

print(f"Video ID: {video_id}")
print(f"Sentence: {sentence}")
print(f"Landmarks shape: {landmarks.shape}") # Expected (T, 96, 3)

# Landmark Indices for the 96 Selected Features
# Order: Face (48), Upper Body (6), Hands (42)
FACE_SLICE = slice(0, 48)
POSE_SLICE = slice(48, 54)
LHAND_SLICE = slice(54, 75)
RHAND_SLICE = slice(75, 96)
Downloads last month
304