One-shot learning of view-invariant object representations in newborn chicks

Can newborn brains perform one-shot learning? To address this question, we reared newborn chicks in strictly controlled environments containing a single view of a single object, then tested their object recognition performance across 24 uniformly-spaced viewpoints. We found that chicks can build view-invariant object representations from a single view of an object: a case of one-shot learning in newborn brains. Chicks can also build the same view-invariant object representation from different views of an object, showing that newborn brains converge on common object representations from different sets of sensory inputs. Finally, by rearing chicks with larger numbers of object views, we found that chicks develop enhanced recognition for familiar views. These results illuminate the earliest stages of object recognition, revealing (1) powerful one-shot learning that builds invariant object representations from the first views of an object and (2) view-based learning that enriches object representations, producing enhanced recognition for familiar views.

(A) Illustration of a controlled-rearing chamber during the Input Phase and Test Phase. During the Input Phase, a single virtual object was presented on one display wall. During the Test Phase, the imprinted object was shown on one display wall and an unfamiliar object was shown on the opposite display wall. (B) The object views presented to chicks during the Input Phase. Chicks were reared with an object moving through either a small (11.25°) or large (360°) viewpoint range. The chicks reared with the small viewpoint range saw either a front or side view of the object. (C) The set of viewpoints presented during the Test Phase, consisting of all possible combinations of 3 elevation rotations and 8 azimuth rotations. The imprinted object and unfamiliar object were presented from the same static viewpoint during each test trial to maximize the image-level similarity between the objects.
Results. (A) The chicks' object recognition performance across the 24 test viewpoints in Experiment 1 (left) and Experiment 2 (right). The red boxes denote familiar views and the black boxes denote novel views. Performance (percent correct) is both color-coded and reported as a numeric value, together with its significance according to a one-tailed t-test (see key for significance values). (B) The chicks successfully recognized their imprinted object across both familiar views (red bars) and novel views (gray bars). The chicks also showed enhanced recognition when the object was presented from familiar views in both Experiment 1 (left) and Experiment 2 (right). The familiar and novel views used in this analysis are denoted by red and black boxes in Panel A, respectively. Error bars indicate standard error. The dashed lines show chance performance. (C) Overall recognition performance by test day. Performance did not improve over the course of the test phase in Experiment 1 (left) or Experiment 2 (right). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
OP Movie 1. training phase with small rotation position 1
OP Movie 2. training phase with small rotation position 2
OP Movie 3. training phase with complete rotation
OP Movie 4. testing phase with distractor object
OP Movie 5. training phase position 2 still
OP Movie 6. training phase position 1 still