pszemraj/inception-trailer-480p

---
license: odc-by
size_categories:
- n<1K
task_categories:
- image-classification
dataset_info:
- config_name: default
features:
- name: image
dtype: image
- name: id
dtype: int64
splits:
- name: train
num_bytes: 87919269
num_examples: 212
download_size: 87746038
dataset_size: 87919269
- config_name: preds-beit-512-shot_types
features:
- name: image
dtype: image
- name: id
dtype: int64
- name: label
dtype: string
- name: score
dtype: float64
splits:
- name: train
num_bytes: 63267510
num_examples: 212
download_size: 63121773
dataset_size: 63267510
- config_name: preds-dinov2-small-film-shot-classifier
features:
- name: image
dtype: image
- name: id
dtype: int64
- name: label
dtype: string
- name: score
dtype: float64
splits:
- name: train
num_bytes: 63267689.0
num_examples: 212
download_size: 63121804
dataset_size: 63267689.0
- config_name: preds-swinv2-base-p4-w12to24-192to384-22kto1k-film-shot-types
features:
- name: image
dtype: image
- name: id
dtype: int64
- name: label
dtype: string
- name: score
dtype: float64
splits:
- name: train
num_bytes: 63267762.0
num_examples: 212
download_size: 63121848
dataset_size: 63267762.0
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- config_name: preds-beit-512-shot_types
data_files:
- split: train
path: preds-beit-512-shot_types/train-*
- config_name: preds-dinov2-small-film-shot-classifier
data_files:
- split: train
path: preds-dinov2-small-film-shot-classifier/train-*
- config_name: preds-swinv2-base-p4-w12to24-192to384-22kto1k-film-shot-types
data_files:
- split: train
path: preds-swinv2-base-p4-w12to24-192to384-22kto1k-film-shot-types/train-*
---

# frames from inception trailer

> i realized that the base video is only in SD quality (lol). will redo as a new dataset with an HD one

These are frames exported from the inception trailer for use in evaluating "real world" predictions by image classifiers.

- Whatever licenses apply to the [original trailer](https://youtu.be/5EiV_HXIIGs?si=fgfSK5lA-3HU_0Qo) extend here.
- Frames are exported in 720p at 2 FPS

## configs

you will find predictions from some finetuned models (+ potentially CLIP??) as separate configs. for the sake of standardization, predictions with score < 0.50 are assigned to `NO_PREDICTION`

原始信息汇总

数据集概述

数据集信息

配置名称：default

• 特征：
  • image：图像数据
  • id：整数标识
• 分割：
  • train：
    • 字节数：87919269
    • 样本数：212
• 下载大小：87746038
• 数据集大小：87919269

配置名称：preds-beit-512-shot_types

• 特征：
  • image：图像数据
  • id：整数标识
  • label：字符串标签
  • score：浮点数得分
• 分割：
  • train：
    • 字节数：63267510
    • 样本数：212
• 下载大小：63121773
• 数据集大小：63267510

配置名称：preds-dinov2-small-film-shot-classifier

• 特征：
  • image：图像数据
  • id：整数标识
  • label：字符串标签
  • score：浮点数得分
• 分割：
  • train：
    • 字节数：63267689.0
    • 样本数：212
• 下载大小：63121804
• 数据集大小：63267689.0

配置名称：preds-swinv2-base-p4-w12to24-192to384-22kto1k-film-shot-types

• 特征：
  • image：图像数据
  • id：整数标识
  • label：字符串标签
  • score：浮点数得分
• 分割：
  • train：
    • 字节数：63267762.0
    • 样本数：212
• 下载大小：63121848
• 数据集大小：63267762.0

数据文件配置

配置名称：default

• 数据文件：
  • train：data/train-*

配置名称：preds-beit-512-shot_types

• 数据文件：
  • train：preds-beit-512-shot_types/train-*

配置名称：preds-dinov2-small-film-shot-classifier

• 数据文件：
  • train：preds-dinov2-small-film-shot-classifier/train-*

配置名称：preds-swinv2-base-p4-w12to24-192to384-22kto1k-film-shot-types

• 数据文件：
  • train：preds-swinv2-base-p4-w12to24-192to384-22kto1k-film-shot-types/train-*