Here are some good points against indirect realism, IMO.
https://philpapers.org/archive/KOORLT.pdf
By giving a causal account of how ideas get formed in the mind as the result of the external world
pouring into us through the senses we can arrive at an epistemological account of how these ideas can be put together in knowledge. Causation here yields justification, or in Rorty’s description, “a quasi-mechanical account of the way in which our immaterial tablets are dented by the material world will help us know what we are entitled to believe” (1979, 143). The history of seventeenth century philosophy forwarded in Mirror has it that the legacy of modern philosophy is a Cartesian-Lockean metaphor in which minds are construed as representing machines whose units of representation are ideas.
...
What Sellars’s “Empiricism and the Philosophy of Mind” helped Rorty to show was that a belief can be shown to be justified (or unjustified) only on the basis of another belief or set of beliefs. A belief cannot be shown to be justified (or not) on the basis of what Sellars mocked in his essay as “the unmoved movers of empirical knowledge” (Sellars 1956, 77). This led Sellars to the point that there is no way to draw a direct link between the supposedly immediate (or non-conceptual) givens of perception and the mediated (or conceptualized) takings of knowledge. For perceptual inputs (e.g., sensations) to be in any way relevant to processes of justification and hence of knowledge they must already be conceptual in form so as to occupy some place in what Sellars called “the logical space of reasons” (1956, 76). Sellars’s claim, upon inspection, is a rather modest one: every conclusion in belief stands in need of
reasons as supporting premises. Modesty, of course, is often a high virtue in philosophy. And in any event, its appearance can be deceptive. In this case, a modest point calls into question the very project of epistemological foundationalism. For what Sellars is suggesting is that as-yet-unconceptualized
perceptual inputs cannot play a determinative role in justificatory practices involving classificatory concepts. The Jamesian “blooming buzzing confusion” of raw sensation may find its way into our experience on occasion but it cannot play any direct justificatory role in so doing.
...
Perceptions are of course conceptually classifiable but not for that reason justifiers of any particular conceptual classification. Every perceptual given is always amenable to a multiplicity of conceptual takings – this is Quine’s thesis of ontological relativity or inscrutability of reference, made memorable in his example of the ‘Gavagai-Rabbit’ translation (Quine 1960, §7ff.). It follows that concepts by themselves do not yield justifications. Concepts are not, merely in virtue of being concepts, justifiers for any other concepts, even though (as Sellars showed) only a conceptually-laden belief can justify a conceptuallyladen belief. Quine’s claim also seems rather modest. But the view it leads to is the radical divorce of epistemology and ontology which follows from the insight that, as Rorty put it, “there is no such thing as direct acquaintance with sensedata or meanings which would give inviolability to reports by virtue of their correspondence to reality, apart from their role in the general scheme of belief”
(1979, 202). Rorty takes Quine to show that perceptions do not enter into us one at a time, but rather as part of complex webs of theory and practice such that any perception is always bundled together with many other perceptions as well as many other beliefs.