From 5978343a73b126c22052c6abfa8f354d3c4581d8 Mon Sep 17 00:00:00 2001 From: Swapnil Sayan Saha Date: Mon, 26 Apr 2021 14:37:09 -0700 Subject: [PATCH] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 6ea0f9a..41c7157 100644 --- a/README.md +++ b/README.md @@ -2,7 +2,7 @@ This repo contains sample code for training deep learning pipelines on multimodal data containing missing and misaligned samples, noisy artifacts and data with variable sampling rates and timing errors, intended for complex event processing. We have benchmarked the pipeline on complex activity recognition using the Cooking Activity Recognition Dataset. [**[Paper](https://www.researchgate.net/publication/341055525_Deep_Convolutional_Bidirectional_LSTM_for_Complex_Activity_Recognition_with_Missing_Data)**] [**[Slides](https://drive.google.com/file/d/1UDM38jvAkwiQRNNMcIAKhnV3b-qJigyF/view)**] [**[Presentation Video](https://www.youtube.com/watch?v=rAgnfvsDK-o)**] -The proposed training pipeline stood 3rd in the **[Cooking Activity Recognition Challenge](https://abc-research.github.io/cook2020/)** out of 78 teams and was awarded the IEEE Lance Stafford Larson Student Award by IEEE Computer Society. +The proposed training pipeline stood 3rd in the **[Cooking Activity Recognition Challenge](https://abc-research.github.io/cook2020/)** out of 78 teams and was awarded the 2020 **[IEEE Lance Stafford Larson Student Award by IEEE Computer Society](https://www.computer.org/volunteering/awards/scholarships/larson). ## Summary Complex activity recognition using multiple on-body sensors is challenging due to missing samples, misaligned data times tamps across sensors, and variations in sampling rates. In this paper, we introduce a robust training pipeline that handles sampling rate variability, missing data, and misaligned data time stamps using intelligent data augmentation techniques. Specifically, we use controlled jitter in window length and add artificial misalignments in data timestamps between sensors, along with masking representations of missing data. We evaluate our pipeline on the Cooking Activity Dataset with Macro and Micro Activities, benchmarking the performance of deep convolutional bidirectional long short-term memory (DCBL) classifier. In our evaluations, DCBL achieves test accuracies of 88% and 72%, respectively, for macro- and micro-activity classifications, exceeding performance over state-of-the-art vanilla activity classifiers.