We propose a weakly supervised temporal action localization algorithm on untrimmed videos using convolutional neural networks. Our algorithm learns from video-level class labels and predicts temporal intervals of human actions with no requirement of temporal localization annotations. We design our network to identify a sparse subset of key segments associated with target actions in a video using an attention module and fuse the key segments through adaptive temporal pooling. Our loss function is comprised of two terms that minimize the video-level action classification error and enforce the sparsity of the segment selection. At inference time, we extract and score temporal proposals using temporal class activations and class-agnostic attentions to estimate the time intervals that correspond to target actions. The proposed algorithm attains state-of-the-art results on the THUMOS14 dataset and outstanding performance on ActivityNet1.3 even with its weak supervision.

No abstract

No abstract

No abstract

The Saigon River, Southern Vietnam, crosses one of the most dynamic developing Megacity in Southeast Asia: Ho Chi Minh City (HCMC). The increased economic, industrial, and domestic developments may affect the environmental quality of water and halieutic resources. In this study, we evaluated the seasonal (dry and wet seasons) biogeochemical state of the Saigon River during two snapshot campaigns conducted along the river basin upstream from HCMC; the Saigon River was characterized by slightly acidic (pH 5.7-7.7) and oxygen-depleted water (dissolved oxygen (DO), 0.36-5.18 mg l). Nutrients (N-NH = 0.01-2.41, N-NO = 0.14-2.72, and P-PO = ~0-0.42 mg l), DOC (2.2-8.0 mg l), POC, and trace metal(oid) (As, Cd, Cr, Cu, Zn, and Hg) concentrations were low showing a good quality of the upstream river. In the urban center area, DO dropped to 0.03 mg l accompanied with a rise of nutrient concentrations (e.g., N-NH, up to 17.7 mg l) likely originating from wastewater discharges. Trace metal concentrations also rose sharply (e.g., Cr and Hg rose up to 10-fold higher) in both water and sediments but remained under the World Health Organization (WHO) and Vietnamese concentration guidelines. In the downstream estuarine area, the intrusion of marine waters diluted water flowing from HCMC, leading water quality to return close to the state observed upstream from HCMC. In general, levels of nutrient and metal contaminations along the Saigon River during both seasons appear moderate regarding to Vietnamese and WHO guidelines although the urban area is highlighted as the major contributor for metal(oid) emissions. Finally, we showed that apart from wastewater and industrial discharges that affect the river quality, metal(oid) partitioning between solid and solution is controlled by the change in water geochemistry along the continuum during both seasons, such as DO (e.g., for As and Cr) and pH (e.g., for Pb) which drives their sorption/dissolution dynamics.