Dataset Preview
Full Screen
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code:   DatasetGenerationCastError
Exception:    DatasetGenerationCastError
Message:      An error occurred while generating the dataset

All the data files must have the same columns, but at some point there are 8 new columns ({'0.151', '0.283', '-5KQ66BBWC4', '0.811', '0.077', '1', '80', '0902'}) and 5 missing columns ({'time_start', 'split', 'label', 'youtube_id', 'time_end'}).

This happened while the csv dataset builder was generating data using

hf://datasets/iejMac/CLIP-Kinetics700/data/annotations/AVA-Kinetics/ava_kinetics_v1_0/ava_train_v2.2.csv (at revision 9e6b82e3134265d63fad9308eb996dbe21b2653c)

Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 2011, in _prepare_split_single
                  writer.write_table(table)
                File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/arrow_writer.py", line 585, in write_table
                  pa_table = table_cast(pa_table, self._schema)
                File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 2302, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 2256, in cast_table_to_schema
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              -5KQ66BBWC4: string
              0902: int64
              0.077: double
              0.151: double
              0.283: double
              0.811: double
              80: double
              1: double
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 1142
              to
              {'label': Value(dtype='string', id=None), 'youtube_id': Value(dtype='string', id=None), 'time_start': Value(dtype='int64', id=None), 'time_end': Value(dtype='int64', id=None), 'split': Value(dtype='string', id=None)}
              because column names don't match
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1316, in compute_config_parquet_and_info_response
                  parquet_operations, partial = stream_convert_to_parquet(
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 909, in stream_convert_to_parquet
                  builder._prepare_split(
                File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 1882, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 2013, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
              datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
              
              All the data files must have the same columns, but at some point there are 8 new columns ({'0.151', '0.283', '-5KQ66BBWC4', '0.811', '0.077', '1', '80', '0902'}) and 5 missing columns ({'time_start', 'split', 'label', 'youtube_id', 'time_end'}).
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/iejMac/CLIP-Kinetics700/data/annotations/AVA-Kinetics/ava_kinetics_v1_0/ava_train_v2.2.csv (at revision 9e6b82e3134265d63fad9308eb996dbe21b2653c)
              
              Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

label
string
youtube_id
string
time_start
int64
time_end
int64
split
string
clay pottery making
---0dWlqevI
19
29
train
news anchoring
---aQ-tA5_A
9
19
train
using bagging machine
---j12rm3WI
14
24
train
javelin throw
--07WQ2iBlw
1
11
train
climbing a rope
--0NTAs-fA0
29
39
train
sipping cup
--0l35AkU34
68
78
train
flipping pancake
--33Lscn6sk
4
14
train
tickling
--3OAstUWtU
45
55
train
watering plants
--3lTx87ebQ
23
33
train
eating spaghetti
--3ouPhoy2A
20
30
train
dribbling basketball
--4-0ihtnBU
58
68
train
calligraphy
--4NLFGNfAs
11
21
train
playing tennis
--56QUhyDQM
185
195
train
brushing floor
--5Tg3gW4s4
8
18
train
chiseling stone
--5kVeU1uco
58
68
train
crossing eyes
--5urtlw6gs
49
59
train
mountain climber (exercise)
--65_UAU7ao
0
10
train
tap dancing
--6q_33gNew
132
142
train
planing wood
--6zxbYq5M4
207
217
train
using inhaler
--71SekUwOA
13
23
train
playing saxophone
--7VUM9MKg4
136
146
train
slapping
--7goKgS4kc
15
25
train
shaking head
--8FQVwWH0M
3
13
train
driving tractor
--9s8lCov-I
2
12
train
delivering mail
--AC9Wionlg
0
10
train
riding a bike
--DiygSPius
52
62
train
calligraphy
--Dt5ovTecs
3
13
train
putting in contact lenses
--EZD1uEVhM
63
73
train
climbing a rope
--EaS9P7ZdQ
13
23
train
shaving head
--FTpoPhxcA
570
580
train
doing jigsaw puzzle
--GEr5-PyTI
0
10
train
playing bass guitar
--GF746y6UM
496
506
train
playing with trains
--GaSxELz-8
39
49
train
digging
--ILYNHl3e4
541
551
train
brushing teeth
--IPbe5ZMCI
2
12
train
washing hair
--JGqNj9Bv0
33
43
train
opening bottle (not wine)
--KkHPoWdW8
25
35
train
finger snapping
--Kr1PZaPDI
1
11
train
crossing eyes
--LrHkjbax8
3
13
train
doing nails
--MW2zcDiNg
197
207
train
making balloon shapes
--Nixnc9DQg
17
27
train
pouring beer
--O2qCO6zBI
25
35
train
preparing salad
--O4p15yejk
351
361
train
playing violin
--PwWfN1Ae4
97
107
train
feeding birds
--PyMoD3_eg
20
30
train
wading through water
--QG-bjOFSA
0
10
train
cutting watermelon
--QfWXBoeA0
125
135
train
petting cat
--Qq8Y0ywHA
2
12
train
playing drums
--RVKBMp_2M
79
89
train
doing aerobics
--RtHwqbEWA
139
149
train
sausage making
--RuMpNXawk
277
287
train
playing ukulele
--SOz3xjWfA
37
47
train
throwing knife
--SSk3r84yw
0
10
train
skiing slalom
--SrV9bGzRA
197
207
train
knitting
--T4dtRsSSg
25
35
train
head stand
--U3pqp_C5g
139
149
train
dining
--UW3CcUqaU
20
30
train
hurdling
--VnA3ztuZg
57
67
train
throwing water balloon
--VtCsmlD18
47
57
train
tiptoeing
--WckmtKBpc
17
27
train
news anchoring
--XQ47BGhdc
46
56
train
opening present
--XR9XONO2U
460
470
train
changing gear in car
--XWWLL8Spk
0
10
train
playing cymbals
--Y25nDn2Wk
60
70
train
chopping meat
--YKMD5LYtw
89
99
train
making paper aeroplanes
--YXAAC0WWA
35
45
train
folding clothes
--ZrYtRjYfs
87
97
train
playing pinball
--_3Wbv-1YY
319
329
train
changing gear in car
--_fOePZDoY
41
51
train
vacuuming floor
--bQb-k_Tjs
9
19
train
shuffling feet
--cIG2WqOf4
20
30
train
recording music
--cLZr4EsEk
133
143
train
historical reenactment
--cOOZbreL4
129
139
train
surfing water
--coBvtS-eQ
57
67
train
playing drums
--d6UCSGoHg
51
61
train
bowling
--dVV4_CSvw
33
43
train
snowkiting
--feYHvtteA
25
35
train
lighting fire
--gMWIQlLvs
122
132
train
eating cake
--gtKHP3Q8E
0
10
train
inflating balloons
--gx7yb1-x0
298
308
train
letting go of balloon
--h55t9J0Xo
7
17
train
calligraphy
--hZSQ7Q7qs
145
155
train
chopping meat
--iGd5Zf7-k
8
18
train
bee keeping
--iIcadFu9c
90
100
train
cutting apple
--ijHZ19K_M
25
35
train
drooling
--ixUOhZodk
56
66
train
gymnastics tumbling
--jD1Yu5ZnQ
37
47
train
archaeological excavation
--jfcGztatc
6
16
train
clapping
--jktjgj81k
5
15
train
giving or receiving award
--kbTDDIiP0
62
72
train
cutting cake
--kyWqclQ24
256
266
train
visiting the zoo
--lmBtwsdBc
0
10
train
marriage proposal
--loSjz82iU
83
93
train
petting animal (not cat)
--lrRHlpK68
218
228
train
egg hunting
--lvipPLwp0
0
10
train
planting trees
--mBL7P45yE
394
404
train
biking through snow
--mI_-gaZLk
18
28
train
playing ping pong
--mLOzFpQoo
7
17
train
brushing teeth
--mQfG6Wu48
28
38
train
dribbling basketball
--mTlk9ommA
48
58
train
End of preview.

Dataset Card for CLIP-Kinetics70

Dataset Description

Dataset Summary

CLIP-Kinetics700 is a compressed version of the Kinetics700 dataset using OpenAI's CLIP model.

The original dataset is ~700 GB making it difficult to use and hold in memory on one machine. By downsampling each video to 1 FPS and encoding the frames using CLIP we we're able to compress the dataset to ~8 GB making it very memory-friendly and easy to use.

Dataset Preprocessing

clip-video-encode is a tool you can use to easily and efficiently compute CLIP embeddings from video frames. We used it to generate the embeddings for this dataset.

Dataset Structure

Data Format

We formatted this as a WebDataset for better data-loading performance when training the models. Each split contains a list of tar files each with 10000 data samples. This format can be read and used easily using the EmbeddingWebDatasetReader from clip-video-encode.

CLIP-Kinetics700
 β”œβ”€β”€ splits.csv
 β”œβ”€β”€ ds_00000.tar
 |     β”œβ”€β”€ vid_00000.npy
 |     β”œβ”€β”€ vid_00000.txt
 |     β”œβ”€β”€ vid_00000.json
 |     β”œβ”€β”€ vid_00001.npy
 |     β”œβ”€β”€ vid_00001.txt
 |     β”œβ”€β”€ vid_00001.json
 |     └── ...
 |     β”œβ”€β”€ vid_10000.npy
 |     β”œβ”€β”€ vid_10000.txt
 |     β”œβ”€β”€ vid_10000.json
 β”œβ”€β”€ ds_00001.tar
 |     β”œβ”€β”€ vid_10001.npy
 |     β”œβ”€β”€ vid_10001.txt
 |     β”œβ”€β”€ vid_10001.json
 β”‚     ...
 ...

Data Fields

  • vid.npy: the numpy array with the per-frame embeddings. Shape -> (n_frames, 512)
  • vid.cap: the "caption" of the video. In this case it is the Kinetics700 label.
  • vid.json: additional metadata - YouTube video ID, start time, end time.

Data Splits

  • Train - 536489 samples | 54 tar's
  • Validation - 33966 samples | 4 tar's
  • Test - 64532 samples | 7 tar's

Dataset Creation

Source Data

Data was sourced from DeepMind's Kinetics700 dataset and downloaded using this convenient repository.

Simple Experiments

Using this repository we evaluate CLIP-Kinetics700 with the following simple methods:

Zero-shot Evaluation

Accuracy
Top-1 0.31
Top-5 0.56
mean(Top1, Top5) 0.44

Linear-probe Evaluation

Accuracy
Top-1 0.41
Top-5 0.65
mean(Top1, Top5) 0.53
Downloads last month
69