The System of Rice Intensification (SRI) has been attributed to improvement in rice production with various attributes being accrued from application of the SRI Principles. The most notable are savings on water use an...The System of Rice Intensification (SRI) has been attributed to improvement in rice production with various attributes being accrued from application of the SRI Principles. The most notable are savings on water use and increase in yield. Alternate Wetting and Drying (AWD) has also paved way for mechanical weed control in paddy fields. One of the major constraints to adoption of SRI is the perceived increased labour input due to the careful transplanting and frequent weed control. This paper evaluates the effect of mechanization on labour input in SRI in comparison to the less mechanized farmer practice. In attempt to reduce drudgery in transplanting under SRI, the drum seeder was used to establish the rice crop by direct seeding. This was then followed by using SRI practices i.e. AWD and mechanical weeding. Direct seeding using a drum seeder was compared to transplanting in both SRI and the common farmer practice. Hand weeding was also evaluated and compared to mechanical weeding. Labour input cost was also compared to the income accrued from the yields. From the study, it was noted that direct seeding using the drum seeder reduced labour input by 97% compared to transplanting. This was possible in that in direct seeding, and there was no nursery preparation and management as in transplanting. The use of a mechanical weeder reduced labour input by 28.3% in relation to hand weeding. Labour input cost for SRI was cheaper (Kshs. 124,080 per hectare) compared to the common farmer practice (Kshs. 139,117.50 per hectare). There was more yield from the SRI practice (2.75 Ton/ha) compared to the common farmer practice (1.88 Ton/ha).展开更多
It is well known that quackgrass is both very aggressive and persistent. In agriculture, many attempts have been made to eliminate this weed without success. Within the context of a sustainable agriculture, mechanical...It is well known that quackgrass is both very aggressive and persistent. In agriculture, many attempts have been made to eliminate this weed without success. Within the context of a sustainable agriculture, mechanical control of quackgrass represents an interesting alternative to chemical means. The use of a ＂C＂ shaped mounted tine cultivator, a rotary cross-harrow, and an ＂S＂ shaped trailed tine cultivator （alone or in combination） to control quackgrass in grassland was investigated. The rate of quackgrass present in each experimental plot was determined before and after the treatments using a one square meter quadrant. Also, the time required for each passage as well as the fuel consumption were computed. Results revealed that the fuel consumption and the time required by the ＂C＂ and ＂S＂ shaped tine cultivators to transport and expose the rhizomes to the sun on the soil surface highly depend on the quality of tillage during the first stubble passage. Also, subsequent tillage with the ＂C＂ shaped tine cultivator after a first passage with a rotary cross-harrow resulted in less fuel consumption and passage time. On the other hand, stubble and subsequent soil tillage with only the ＂C＂ shaped tine cultivator is the least successful method.展开更多
针对卷积神经网络模型巨大的参数量和计算量导致其实际应用时难度较大的问题,提出了一种基于注意力机制与动态稀疏约束的模型压缩方法。该算法首先借助SENet(Squeeze and excitation networks,SENet)模块(可称为SE模块)评估出网络中各...针对卷积神经网络模型巨大的参数量和计算量导致其实际应用时难度较大的问题,提出了一种基于注意力机制与动态稀疏约束的模型压缩方法。该算法首先借助SENet(Squeeze and excitation networks,SENet)模块(可称为SE模块)评估出网络中各个通道的重要性,并施加稀疏正则化;然后提出一种网络稀疏度的自适应惩罚权重设计方法,根据模型学习效果,动态调整权重,将其添加到最终的训练目标上,实现模型动态压缩。最后,通过实验验证所提出的模型压缩方法,在经典的多分类数据集CIFAR 10上进行实验,证明了本文所提出的基于注意力机制与动态稀疏约束的模型压缩方法可降低网络的冗余度,使网络模型参数量减少43.97%,计算量减少82.94%,而分类准确率只比原始VGG16模型下降0.04个百分点。随后又将提出的模型压缩方法应用到杂草检测任务中,在甜菜与杂草数据集上进行实验,实验结果表明,剪枝模型相较于未剪枝模型的模型参数量减少41.26%,计算量减少45.77%,而平均检测精度均值只减少0.91个百分点,证明了该方法在杂草检测方面效果较好。展开更多
文摘The System of Rice Intensification (SRI) has been attributed to improvement in rice production with various attributes being accrued from application of the SRI Principles. The most notable are savings on water use and increase in yield. Alternate Wetting and Drying (AWD) has also paved way for mechanical weed control in paddy fields. One of the major constraints to adoption of SRI is the perceived increased labour input due to the careful transplanting and frequent weed control. This paper evaluates the effect of mechanization on labour input in SRI in comparison to the less mechanized farmer practice. In attempt to reduce drudgery in transplanting under SRI, the drum seeder was used to establish the rice crop by direct seeding. This was then followed by using SRI practices i.e. AWD and mechanical weeding. Direct seeding using a drum seeder was compared to transplanting in both SRI and the common farmer practice. Hand weeding was also evaluated and compared to mechanical weeding. Labour input cost was also compared to the income accrued from the yields. From the study, it was noted that direct seeding using the drum seeder reduced labour input by 97% compared to transplanting. This was possible in that in direct seeding, and there was no nursery preparation and management as in transplanting. The use of a mechanical weeder reduced labour input by 28.3% in relation to hand weeding. Labour input cost for SRI was cheaper (Kshs. 124,080 per hectare) compared to the common farmer practice (Kshs. 139,117.50 per hectare). There was more yield from the SRI practice (2.75 Ton/ha) compared to the common farmer practice (1.88 Ton/ha).
文摘It is well known that quackgrass is both very aggressive and persistent. In agriculture, many attempts have been made to eliminate this weed without success. Within the context of a sustainable agriculture, mechanical control of quackgrass represents an interesting alternative to chemical means. The use of a ＂C＂ shaped mounted tine cultivator, a rotary cross-harrow, and an ＂S＂ shaped trailed tine cultivator （alone or in combination） to control quackgrass in grassland was investigated. The rate of quackgrass present in each experimental plot was determined before and after the treatments using a one square meter quadrant. Also, the time required for each passage as well as the fuel consumption were computed. Results revealed that the fuel consumption and the time required by the ＂C＂ and ＂S＂ shaped tine cultivators to transport and expose the rhizomes to the sun on the soil surface highly depend on the quality of tillage during the first stubble passage. Also, subsequent tillage with the ＂C＂ shaped tine cultivator after a first passage with a rotary cross-harrow resulted in less fuel consumption and passage time. On the other hand, stubble and subsequent soil tillage with only the ＂C＂ shaped tine cultivator is the least successful method.
文摘针对卷积神经网络模型巨大的参数量和计算量导致其实际应用时难度较大的问题,提出了一种基于注意力机制与动态稀疏约束的模型压缩方法。该算法首先借助SENet(Squeeze and excitation networks,SENet)模块(可称为SE模块)评估出网络中各个通道的重要性,并施加稀疏正则化;然后提出一种网络稀疏度的自适应惩罚权重设计方法,根据模型学习效果,动态调整权重,将其添加到最终的训练目标上,实现模型动态压缩。最后,通过实验验证所提出的模型压缩方法,在经典的多分类数据集CIFAR 10上进行实验,证明了本文所提出的基于注意力机制与动态稀疏约束的模型压缩方法可降低网络的冗余度,使网络模型参数量减少43.97%,计算量减少82.94%,而分类准确率只比原始VGG16模型下降0.04个百分点。随后又将提出的模型压缩方法应用到杂草检测任务中,在甜菜与杂草数据集上进行实验,实验结果表明,剪枝模型相较于未剪枝模型的模型参数量减少41.26%,计算量减少45.77%,而平均检测精度均值只减少0.91个百分点,证明了该方法在杂草检测方面效果较好。
