Apparatus, method, device and medium for label-balanced calibration in post-training quantization of dnn

Patent metadata

Field	Value
Publication number	US-2025077861-A1
Application number	US-202118574995-A
Country	US
Kind code	A1
Filing date	Nov 3, 2021
Priority date	Nov 3, 2021
Publication date	Mar 6, 2025
Grant date	—

Abstract

The disclosure provides an apparatus, method, device and medium for label-balanced calibration in post-training quantization of DNNs. An apparatus includes interface circuitry configured to receive a training dataset and processor circuitry coupled to the interface circuitry. The processor circuitry is configured to generate a small ground truth dataset by selecting images with a ground truth number of 1 from the training dataset; generate a calibration dataset randomly from the training dataset; if any image in the calibration dataset has the ground truth number of 1, remove the image from the small ground truth dataset; generate a label balanced calibration dataset by replacing an image with a ground truth number greater than a preset threshold in the calibration dataset with a replacing image selected randomly from the small ground truth dataset; and perform calibration using the label balanced calibration dataset in post-training quantization. Other embodiments are disclosed and claimed.

First claim

1 - 24 . (canceled)

25 . An apparatus, comprising:
interface circuitry configured to receive a training dataset; and processor circuitry coupled to the interface circuitry, the processor circuitry being configured to:
generate a small ground truth dataset by selecting images with a ground truth number of 1 from the training dataset;
generate a calibration dataset randomly from the training dataset;
if any image in the calibration dataset has the ground truth number of 1, remove the image from the small ground truth dataset;
generate a label balanced calibration dataset by replacing an image with a ground truth number greater than a preset threshold in the calibration dataset with a replacing image selected randomly from the small ground truth dataset; and
perform calibration using the label balanced calibration dataset in post-training quantization of a deep neural network (DNN).

26 . The apparatus of claim 25 , wherein the processor circuitry is configured to generate the label balanced calibration dataset by,
for each image in the calibration dataset,
appending the image to the label balanced calibration dataset, under a condition that a ground truth number of the image is not greater than the preset threshold; or
under a condition that the ground truth number of the image is greater than the preset threshold,
selecting randomly the replacing image for the image from the small ground truth dataset,
appending the replacing image to the label balanced calibration dataset, and
removing the replacing image from the small ground truth dataset.

27 . The apparatus of claim 25 , wherein the preset threshold is 5.

28 . The apparatus of claim 25 , wherein the DNN comprises a multi-label DNN for object detection or instance segmentation.

29 . The apparatus of claim 25 , wherein the training dataset comprises a Common Objects in Context (COCO) dataset.

30 . The apparatus of claim 25 , wherein the post-training quantization reduces precision of parameters of the DNN from float 32 bits to integer 8 bits.

31 . A method, comprising:
generating a small ground truth dataset by selecting images with a ground truth number of 1 from a training dataset; generating a calibration dataset randomly from the training dataset; if any image in the calibration dataset has the ground truth number of 1, removing the image from the small ground truth dataset; generating a label balanced calibration dataset by replacing an image with a ground truth number greater than a preset threshold in the calibration dataset with a replacing image selected randomly from the small ground truth dataset; and performing calibration using the label balanced calibration dataset in post-training quantization of a deep neural network (DNN).

32 . The method of claim 31 , wherein generating the label balanced calibration dataset comprises:
for each image in the calibration dataset,
appending the image to the label balanced calibration dataset, under a condition that a ground truth number of the image is not greater than the preset threshold; or
under a condition that the ground truth number of the image is greater than the preset threshold,
selecting randomly the replacing image for the image from the small ground truth dataset,
appending the replacing image to the label balanced calibration dataset, and
removing the replacing image from the small ground truth dataset.

33 . The method of claim 31 , wherein the preset threshold is 5.

34 . The method of claim 31 , wherein the DNN comprises a multi-label DNN for object detection or instance segmentation.

35 . The method of claim 31 , wherein the training dataset comprises a Common Objects in Context (COCO) dataset.

36 . The method of claim 31 , wherein the post-training quantization reduces precision of parameters of the DNN from float 32 bits to integer 8 bits.

37 . A non-transitory machine readable storage medium having instructions stored thereon, which when executed by a machine, cause the machine to perform operations, comprising:
generating a small ground truth dataset by selecting images with a ground truth number of 1 from a training dataset; generating a calibration dataset randomly from the training dataset; if any image in the calibration dataset has the ground truth number of 1, removing the image from the small ground truth dataset; generating a label balanced calibration dataset by replacing an image with a ground truth number greater than a preset threshold in the calibration dataset with a replacing image selected randomly from the small ground truth dataset; and performing calibration using the label balanced calibration dataset in post-training quantization of a deep neural network (DNN).

38 . The non-transitory machine readable storage medium of claim 37 , wherein generating the label balanced calibration dataset comprises:
for each image in the calibration dataset,
appending the image to the label balanced calibration dataset, under a condition that a ground truth number of the image is not greater than the preset threshold; or
under a condition that the ground truth number of the image is greater than the preset threshold,
selecting randomly the replacing image for the image from the small ground truth dataset,
appending the replacing image to the label balanced calibration dataset, and
removing the replacing image from the small ground truth dataset.

39 . The non-transitory machine readable storage medium of claim 37 , wherein the preset threshold is 5.

40 . The non-transitory machine readable storage medium of claim 37 , wherein the DNN comprises a multi-label DNN for object detection or instance segmentation.

41 . The non-transitory machine readable storage medium of claim 37 , wherein the training dataset comprises a Common Objects in Context (COCO) dataset.

42 . The non-transitory machine readable storage medium of claim 37 , wherein the post-training quantization reduces precision of parameters of the DNN from float 32 bits to integer 8 bits.

Assignees

• Intel Corp

Inventors

• Shen Haihao
• Tian Feng
• Chen Xi
• Abidi Huma
• Zhou Yuwen

Classifications

• G06N3/0495 Primary

  Quantised networks; Sparse networks; Compressed networks · CPC title

• G06V10/774

  Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06V10/776

  Validation; Performance evaluation · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title