Download

Contact

Delegation service

E-mail

Annual Report: CIRAD 2007

Oscar and Oswald, computer-assisted weed identification tools

Accurate plant identifications are generally difficult for non-botanists—standard flora manuals are often hard to follow and identification keys are ill adapted to agricultural constraints. CIRAD has tried to overcome these drawbacks by developing software tools to facilitate plant identification graphically through composite drawings. Two applications have just been implemented to facilitate Asian weed identification.

A page of the Oswald software © CIRAD

The number of flora specialists is drastically declining. This is unfortunate since they have never been in such high demand, especially for studies on biodiversity and agriculture, where it is essential to have a clear picture of the plants to be controlled. Moreover, non-specialists are hampered by many problems when attempting to identify weeds according to conventional flora manuals: identifications that are primarily based on the flowers, whereas weeding should be done before flowering; a dichotomous key system that does not allow errors and has a compulsory question choice and order; and an abstruse terminology.

Software tools that enable users to identify a plant graphically through composite drawings have been developed to overcome these drawbacks. These tools have several advantages. They are based only on drawings, without any technical terminology, and the user is given a choice concerning the characters to describe. Incomplete specimens may be identified and observation errors are tolerated. At each identification step, a likelihood probability is calculated for each target species, which are sorted in decreasing order of likelihood. The user has constant access to photographs, descriptive texts and botanical drawings of the species, and the texts also provide links to illustrated definitions of technical terms. Finally, these software tools indicate the most relevant characters, so as to be able to quickly differentiate species, and highlight character misinterpretations when a species has been identified with a probability of under 100%.

These software programs can be used without any special computer skills or botanical knowledge. Buttons on the screen provide ready access to the different choices by a simple click. All of the technical terms can be replaced by drawings and information displayed on the screen may be printed.

The species identification software IDAO (Identification assistée par ordinateur) was designed by CIRAD to fulfil this need. Two downloadable versions are available, one that was developed with open source software and the other in SVG format. Two other software packages have been developed for the identification of Asian weed species: Open Source Simple Computer for Agriculture in Rural Areas (OSCAR) for the identification of weeds on the Indo-Gangetic Plain; and Open Source for Weed Assessment in Lowland Paddy Fields (OSWALD) for weeds in Cambodian and Laotian rice fields.

OSCAR assists users in identifying and controlling weeds on plains in Pakistan, Nepal, India and Bangladesh. This software can be used on a Simputer, PC or online. It is highly intuitive and fast reference cards can be displayed for each weed species, while also describing the most efficient control methods, in English, Hindi, Bengali and Urdu. OSWALD is a similar software tool that was developed for the identification of weeds in Cambodian and Laotian rice fields. The reference cards are in Khmer and Laotian, as well as English and French. It enables users to identify 113 major weeds. An extension is being prepared for Vietnam.

Contact

Pierre Grard, e-mail , Botany and Computational Plant Architecture (UMR AMAP)

Partners

  • Institut français de Pondichéry (IFP, India)
  • National Agriculture and Forest Research Institute (NAFRI, Laos)
  • Rice-Wheat Consortium for Indo-Gangetic Plains (RWC, South Asia)
  • Leiden University (Netherlands)
  • Wageningen University (Netherlands)
  • Université royale d’agriculture (Cambodgia).

Oscar and Oswald benefited from European Commission funding (Asia IT&C programme).

For further information

The Oscar project was awarded, in September 2007, a prize by the Manthan Foundation, under the auspices of the World Summit Award, in the category “India’s best e-content for development”