<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-0608">
  <Title>Why can't Jos'e read? The problem of learning semantic associations in a robot environment</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5 Discussion and conclusion
</SectionTitle>
    <Paragraph position="0"> Our experiments suggest that we can eliminate the costly step of segmentation without incurring a penalty to the object recognition task. This realisation allows us to remove the main computational bottleneck and pursue real-time learning in a mobile robot setting. Moreover, by introducing spatial relationships into the model, we maintain a degree of consistency between individual patch annotations. We can consider this to be an early form of segmentation that takes advantage of both high-level and low-level information. Thus, we are solving both the segmentation and recognition problems simultaneously.</Paragraph>
    <Paragraph position="1"> However, we emphasize the need for further investigation to pin down the role of segmentation in the image translation process.</Paragraph>
    <Paragraph position="2"> Our translation model is disposed to predicting certain words better than others. However, at this point we cannot make make any strong conclusions as to why certain words easy to classify (e.g. cabinets), while others are difficult (e.g. filers). From Figure 5, it appears to be the case that words that occur frequently and possess a consistent set of features tend to be more easily classified.</Paragraph>
    <Paragraph position="3"> Initially, we were doubtful that spatial context in the model would improve results given that the robot roams in a fairly homogeneous environment. This contrasts with experiments on the Corel data sets (Carbonetto and de Freitas, 2003), whereby the photographs were captured from a wide variety of settings. However, the experiments on the robomedia data demonstrate that there is something to be gained by introducing inter-alignment dependencies in the model, even in environments with relatively noisy and unreliable data.</Paragraph>
    <Paragraph position="4"> Generic object recognition in the context of robotics is a challenging task. Standard low-level features such as colour and texture are particularly ineffective in a laboratory environment. For example, chairs can come in a variety of shapes and colours, and &amp;quot;wall&amp;quot; refers to a vertical surface that has virtually no relation to colour, texture and position. Moreover, it is much more difficult to delineate specific concepts in a scene, even for humans -does a table include the legs, and where does one draw the line between shelves, drawers, cabinets and the objects contained in them? (This explains why many of the manually-annotated patches in Figures 6 and 8 are left empty.) Object recognition on the Corel data set is comparatively easy because the photos are captured to artificially delineate specific concepts. Colour and texture tend to be more informative in natural scenes.</Paragraph>
    <Paragraph position="5"> In order to tackle concerns mentioned above, one approach would be to construct a more sophisticated representation of objects. A more realistic alternative would be to reinforce our representation with high-level features, including more complex spatial relations.</Paragraph>
    <Paragraph position="6"> One important criterion we did not address explicitly is on-line learning. Presently, we train our models assuming that all the images are collected at one time. Research shows that porting batch learning to an on-line process using EM does not pose significant challenges (Smith and Makov, 1978; Sato and Ishii, 2000; Brochu et al., 2003). With the discussion presented in this paper in mind, real-time interactive learning of semantic associations in Jos'e's environment is very much within reach.</Paragraph>
  </Section>
class="xml-element"></Paper>