https://github.com/verisilicon/acuity-dataset

Acuity Dataset scripts and examples
https://github.com/verisilicon/acuity-dataset

acuity deep-learning neural-network

Last synced: 11 months ago
JSON representation

Acuity Dataset scripts and examples

• Host: GitHub
• URL: https://github.com/verisilicon/acuity-dataset
• Owner: VeriSilicon
• Created: 2017-12-14T21:13:21.000Z (over 8 years ago)
• Default Branch: master
• Last Pushed: 2019-01-29T06:17:14.000Z (over 7 years ago)
• Last Synced: 2025-04-30T07:47:06.471Z (about 1 year ago)
• Topics: acuity, deep-learning, neural-network
• Language: Python
• Homepage:
• Size: 69 MB
• Stars: 6
• Watchers: 1
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# ACUITY Dataset Examples

This package contains information about the database format used by ACUITY

for storing Training and Testing images. Sample scripts for ImageNet and 

PASCAL VOC are also provided. 

## ACUITY supported dataset

### SQLite Database format

ImageNet Example (SQlite)

   

> See ReadMe.md in ImageNet folder for details

   

PASCAL Example (SQlite)

   

> See ReadMe.md in PASCAL folder for details

MNIST and INRIA Example (SQlite)

> Pre-built Dataset for Mnist and INRIA

### Text format

ImageNet Exmaple (Text)

> ImageNet/dataset.txt contains example of the text dataset format

ILSVRC2012_VAL_SET (Text)

> ILSVRC2012_VAL_SET/dataset.txt contains a pre-built small validation set of images

  from ILSVRC2012 ImageNet

## SQLite Database Format

   Default.dsx is an empty ACUITY SQlite dataset file

    # Use sqlitebrowser to view Dataset schema and content

    sqlitebrowser Default.dsx

    # Tables

    There are 5 tables in the Dataset

    "files" Table

    This table contains the path to the resource files (typically Images) in the

    Database.

        -id: Unique key identifying the file

        -name: File name without extension

        -ext: File extension

    "groups" Table

    This table contains the group entry - "Training" and "Testing"

        - id-name: 1-Training, 2-Testing

    "labels" Table

    This table contains the labels or classes for associated with the dataset. For

    example, ILSVRC dataset contains 1000 classes with each class identifying a

    particular type of objects 

        - id: Unique key identifying the label

        - name: String name for the label

        - description: String description

    "boxed_annotations" Table

    This is the master table which contains information about all the images in the

    dataset using the metadata in the other tables. 

    

        - id: Unique key identifying this annotation

        - label_id: label id associated with this annotation

        - group_id: Identify whether this annotation is used for Training or Testing

        - file_id: The image file for this annotation

        - x/y/w/h: ROI window from the image where this annotation is located

    "validation_results" Table

    This table stores the validation result from running Validation of the network

        - id: Unique key identifying the result

        - session_timestamp: Time Stamp for the inference

        - file_id: Image file used for this inference

        - boxed_annotation: id key into the box_annotation table

        - expected_label/calculated_label

        - score: Top1 score of the inference

        - result_vector: Values from the output vector of the network

 

## Text Dataset format

    

    Text Dataset format is a comma separated text file with each line containing

    the file location and label.

```shell

 ./data/ILSVRC2012_val_00000347.JPEG, 472, 0 0 28 28, validate # Use for validate only, crop pad "0,0,28,28", lable is 472

 ./data/ILSVRC2012_val_00000355.JPEG, 260, 1 1 29 29, train    # Use for traine only, crop pad "1,1,29,29", lable is 260

 ./data/ILSVRC2012_val_00000361.JPEG, 300, , train             # Use for train only, no crop pad, label is 300

```