• Home
  • History
  • Annotate
Name
Date
Size
#Lines
LOC

..--

coco_object_detection/23-Nov-2023-773609

imagenet_image_classification/23-Nov-2023-565438

inference_diff/23-Nov-2023-297222

BUILDD23-Nov-20231 KiB3733

README.mdD23-Nov-20232.2 KiB4630

build_def.bzlD23-Nov-2023510 1512

task_executor.ccD23-Nov-20231.7 KiB4727

task_executor.hD23-Nov-20231.9 KiB5021

task_executor_main.ccD23-Nov-20231.3 KiB3316

README.md

1# TFLite Model Task Evaluation
2
3This page describes how you can check the accuracy of quantized models to verify
4that any degradation in accuracy is within acceptable limits.
5
6## Accuracy & correctness
7
8TensorFlow Lite has two types of tooling to measure how accurately a delegate
9behaves for a given model: Task-Based and Task-Agnostic.
10
11**Task-Based Evaluation** TFLite has two tools to evaluate correctness on two
12image-based tasks: - [ILSVRC 2012](http://image-net.org/challenges/LSVRC/2012/)
13(Image Classification) with top-K accuracy -
14[COCO Object Detection](https://cocodataset.org/#detection-2020) (w/ bounding
15boxes) with mean Average Precision (mAP)
16
17**Task-Agnostic Evaluation** For tasks where there isn't an established
18on-device evaluation tool, or if you are experimenting with custom models,
19TensorFlow Lite has the Inference Diff tool.
20
21## Tools
22
23There are three different binaries which are supported. A brief description of
24each is provided below.
25
26### [Inference Diff Tool](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/tools/evaluation/tasks/inference_diff#inference-diff-tool)
27
28This binary compares TensorFlow Lite execution in single-threaded CPU inference
29and user-defined inference.
30
31### [Image Classification Evaluation](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/tools/evaluation/tasks/imagenet_image_classification#image-classification-evaluation-based-on-ilsvrc-2012-task)
32
33This binary evaluates TensorFlow Lite models trained for the
34[ILSVRC 2012 image classification task.](http://www.image-net.org/challenges/LSVRC/2012/)
35
36### [Object Detection Evaluation](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/tools/evaluation/tasks/coco_object_detection#object-detection-evaluation-using-the-2014-coco-minival-dataset)
37
38This binary evaluates TensorFlow Lite models trained for the bounding box-based
39[COCO Object Detection](https://cocodataset.org/#detection-eval) task.
40
41********************************************************************************
42
43For more information visit the TensorFlow Lite guide on
44[Accuracy & correctness](https://www.tensorflow.org/lite/performance/delegates#accuracy_correctness)
45page.
46