/pdf/landslide-inventory-maps-new-tools-for-an-old-problem-wkxin2jqxg.pdf

Landslide inventory maps: New tools for an old problem

Remote sensing of landslides: An analysis of the potential contribution to geo-spatial systems for hazard assessment in mountainous environments

Semi-automatic recognition and mapping of rainfall induced shallow landslides using optical satellite images

IEEE International Conference on Fuzzy Systems

Catastrophic debris flows on 13 August 2010 in the Qingping area, southwestern China: The combined effects of a strong earthquake and subsequent rainstorms

Assessing debris-flow hazard in a watershed in Taiwan

The essential work of sewer rehabilitation is a sewer inspection through diagnoses of sewer pipe defects. At present, image processing and artificial intelligence techniques have been used to develop diagnostic systems to assist engineers in interpreting sewer pipe defects on CCTV images to overcome human's fatigue and subjectivity, and time-consumption. Based on the segmented morphologies on images, the diagnostic systems were proposed to diagnose sewer pipe defects. However, the environmental influence and image noise hamper the efficiency of automatic diagnosis. For example, the central area of a CCTV image, where is always darker than the surrounding due to the vanishing light and slight reflectance, causes a difficulty to segment correct morphologies. In this paper, a novel approach of morphological segmentation based on edge detection (MSED) is presented and applied to identify the morphology representatives for the sewer pipe defects on CCTV images. Compared with the performances of the opening top-hat operation, which is a popular morphological segmentation approach, MSED can generate better segmentation results. As long as the proper morphologies of sewer pipe defects on CCTV images can be segmented, the morphological features, including area, ratio of major axis length to minor axis length, and eccentricity, can be measured to effectively diagnose sewer pipe defects.

Morphological segmentation based on edge detection for sewer pipe defects on CCTV images

Multiobjective optimization using nondominated sorting genetic algorithm-II for allocation of energy conservation and renewable energy facilities in a campus

Risk assessment of debris flows in Songhe Stream, Taiwan

Large language models (LLMs) show excellent performance but are compute- and memory-intensive. Quantization can reduce memory and accelerate inference. However, existing methods cannot maintain accuracy and hardware efficiency at the same time. We propose SmoothQuant, a training-free, accuracy-preserving, and general-purpose post-training quantization (PTQ) solution to enable 8-bit weight, 8-bit activation (W8A8) quantization for LLMs. Based on the fact that weights are easy to quantize while activations are not, SmoothQuant smooths the activation outliers by offline migrating the quantization difficulty from activations to weights with a mathematically equivalent transformation. SmoothQuant enables an INT8 quantization of both weights and activations for all the matrix multiplications in LLMs, including OPT, BLOOM, GLM, MT-NLG, and LLaMA family. We demonstrate up to 1.56x speedup and 2x memory reduction for LLMs with negligible loss in accuracy. SmoothQuant enables serving 530B LLM within a single node. Our work offers a turn-key solution that reduces hardware costs and democratizes LLMs. Code is available at https://github.com/mit-han-lab/smoothquant.

Ji-Yuan Lin

Papers

Assessing debris-flow hazard in a watershed in Taiwan

Morphological segmentation based on edge detection for sewer pipe defects on CCTV images

Multiobjective optimization using nondominated sorting genetic algorithm-II for allocation of energy conservation and renewable energy facilities in a campus

Risk assessment of debris flows in Songhe Stream, Taiwan

SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models