All Products
Search
Document Center

Platform For AI:LVM-NSFW Filter (DLC)

Last Updated:Jun 03, 2026

Filters videos with high NSFW (Not Safe For Work) scores from your dataset. Only MP4 videos are supported.

Supported computing resources

Deep Learning Containers (DLC)

How it works

Samples frames from each video and calculates NSFW scores per frame. Videos exceeding the score threshold are filtered out. Typically used to curate training data for video generation models.

Inputs and outputs

Input ports

  • Use the Read File Data component to read training data from OSS.

  • Configure OSS Data Path to select the OSS directory containing video data or the video metadata file. See the parameter table below.

  • Any LVM Data Processing (DLC) component can serve as input.

Output port

The filtering results. See the parameter table below for output details.

Configure the component

Configure the LVM-NSFW Filter (DLC) component in Machine Learning Designer with the following parameters.

Tab

Parameter

Required

Description

Default value

Field Settings

Video Data OSS Path

No

If no upstream component exists on first run, manually select the OSS directory storing video data. The component generates a video_meta.jsonl metadata file in the OSS bucket root directory. For subsequent runs, select video_meta.jsonl directly.

No default value

Output File OSS Path

Yes

The OSS directory for filtering results. Output files:

  • {name}.jsonl: the output file. Specify the file name with Output Filename.

  • {name}_stats.jsonl: the stats file.

  • dj_run_yaml.yaml: the runtime configuration file.

No default value

Output Filename

Yes

File name for the filtering results.

result.jsonl

Parameter Settings

NSFW Score Threshold

Yes

The maximum NSFW score.

0.5

Number of Sampled Frames

Yes

Number of frames to sample. Frames are evenly distributed across the video duration.

1

Execution Tuning

Number of Processes

Yes

The number of processes.

4

Select Resource Group

Public Resource Group

No

Instance type (CPU or GPU), number of instances, and VPC.

No default value

Dedicated resource group

No

vCPUs, memory, shared memory, GPUs, and number of instances.

No default value

Maximum Running Duration (seconds)

No

Maximum runtime in seconds. The job terminates if this limit is exceeded.

No default value